Replacing Big SaaS - How to cut the Google, Apple, Dropbox, Microsoft, ... cords

With a Prism and Snowden inspired kick in the backside I finally got around to establishing some autonomy from the Big Boys with respect to email, contacts, calendar, network storage/sync and other common personal use SaaSs.  No rocket science here, just a consolidation of lots of "which one is best for me" research, "follow the tutorial" efforts and Google and log file problems/solutions to explain how to install, configure and maintain the types of services you get "for free" from Google, Apple, Dropbox and the rest.

This article is an overview of how to accomplish replacing the important Big SaaS, it is not a detailed step-by-step with every command listed.  I reference a number of other web pages and tutorials to help with the harder parts.

Overview

Here is a basic overview of the substitutions:

Service	Before	After
Hosting and OS	Google, Apple, Microsoft, Yahoo, ...	Digital Ocean "Droplets" Linux
Email	Google, Apple, Microsoft, Yahoo, ...	postfix, dovecot
Contacts	Google, Apple	davical
Calendar	Google, Apple	davical
Network storage and sync	Dropbox, Copy, Google Drive	ownCloud

The aspirational criteria I had for the substitutions were:

• Open source
• Supported with apt-get or similar installer with an up-to-date stable version available
• At least some recent community activity and support
• Positive reviews, particularly as versus their popular commercial alternatives
• Free or close to it
• Targeted solutions, not one package that is providing many services (e.g., MS Exchange vs Postfix)

It's also important to keep in mind that these solutions generally won't be as good as their popular commercial alternatives where armies of developers and systems administrators support them and taking advantage of big economies of scale and underpricing.  To take this path you're going to forfeit convenience, better usability, rock solid systems and uptime, macro level security, and "free" pricing for greater privacy and control.

Lastly, there are many more areas that could be substituted and I've not done or written these up yet - I note at least some of them at the bottom of article.

What's Required From You

You have to be able to do the following to get this working:

• Basic Unix shell commands and configuration file editing
• Willingness to read various tutorials and how-tos and be able to google for the rest
• Willingness to pay $5 per month for hosting and another $1 per month for backups
• Accept having a total data footprint of 15GB or less (or be willing to pay for more storage)
• A basic understanding of SSL certificates is useful

1. Create an SSH key

Follow Digital Ocean's tutorial to create your own key.

2. Have a domain name ready to use

There are many companies that offer domain registration.

3. Hosting

Set up an account with Digital Ocean (digitalocean.com).  Their basic IaaS virtual server ("Droplet") is cheap, plenty performant for our uses here and their management and provisioning interface is pleasantly usable.

Buy the cheapest cheapest droplet at $5 per month (1 CPU, 512MB RAM, 20GB Disk, 1TB transfer).  This will provide plenty of horsepower and space for the average user.

You might select "Amsterdam" as your region if you thought that might provide a safer environment for your data as opposed to hosting that is based in the USA (Digital Ocean's other sites are in New York and San Francisco).

Select OS "Ubuntu 12.04 x64".  You could probably safely use the newer versions, I've just not moved up to them yet.

Install the SSH certificate you created in step 1.

Enable "VirtIO" if you want.  Whatever it is.

After your new virtual server is created, activate automatic backups for it.  They may only be taken about once per week but they're a bargain at $1 per month.

Set up your new domain name to point to your new droplet IP address.  Digital Ocean's DNS interface is easier than godaddy's.  Configure your domain to use Digital Ocean's DNS.

NOTE: The only thing I don't like about Digital Ocean for hosting is there is no apparent way to cost effectively scale just disk size.  I'd like to keep the memory and CPU of the smallest instance but then easily scale up disk space.  Replacing network storage and big IMAP email archives will exceed the 20GB limit for "power" users.  There are plenty of other providers and some allow a low-performance-high-disk-space specification.  However, among the usual suspects like Amazon and Rackspace along with a number of others I found googling around, I didn't find any in the same price range as Digital Ocean.  Maybe Digital Ocean will add the feature of cost effectively adding disk space only in the future.

4. Basics

Verify you can log in as root using ssh and the ssh certificate you created.

Restrict root login to only allow certificate based logins.

Create a new user that you'll use to do most work from here forward.

Enable new user for sudo use.

Install zsh (or your preferred shell if its not already present) and make it your default shell.  Update your login shell preferences.

Create/deploy another ssh certificate for the new user you've created.

Install ntp.

Install iptables as your firewall.  Digital Ocean has a good tutorial. 

5. Supporting applications

Before we get to the applications we want, we have to install their supporting applications.

Install postgres - used by davical

Install MySQL - used by ownCloud

Install Apache and PHP - used by almost everything

Install phppgadmin - used to administer the Postgres / davical database

Install phpmyadmin - used to administer the MySQL / ownCloud database

6. Create a free SSL Certificate and install it

The certificate will be used by a number of services we install.

Use this tutorial at arstechnica to create a free Class 1 SSL certificate with startssl.com.

Tips:

• startssl.com creates an S/MIME and authentication certificate and automatically installs in your browser.  You might want to save the authentication certificate someplace secure.
• Certificate only good for one year - just remember you need to renew it each year (all your services dependant on a valid SSL cert will stop working when cert expires)

7. Email

Note: I don't typically use webmail, so I didn't bother installing a webmail service.

Install postfix - see Digital Ocean tutorial

Install dovecot - also see Digital Ocean tutorial, my user comments on dovecot

Update DNS MX record.

Adjust iptables firewall settings - see Digital Ocean tutorial

Tips:

• I found "apt-get install mail-stack-delivery" did the heavy lifting for me here.
• Make sure you un/comment out exactly what you want in /etc/postfix/master.cf
• Increased value of mail_max_userip_connections from 10 to 30 in /etc/dovecot/conf.d/01-mail-stack-delivery.conf due to an IMAP error limit popping up in OS X mail.
• Digital Ocean has subsequently created a tutorial for iRedMail - looks easier to set up and includes a webmail interface

Note: not added in spam filtering yet.

8. Contacts and Calendar

Install davical.

I looked at and discounted the following:

• calendarserver - depends on extended file attributes; apt-get exists but doesn't appear to be maintained
• radicale - no backoffice, feels too barebones
• baikai - No apt-get; synology's choice for their sync app
• ownCloud - ownCloud already looks bloated

9. Network storage and sync

Install ownCloud.

The goal here is secure and pervasively available files.  Like Dropbox and the paid version of BoxCryptor - both of which are closed source and therefore non-starters with my stated criteria.

You can create an encrypted filesystem on your main OS, ideally once that can be used by several OSs and place the system in ownCloud network synced storage.  When choosing an filesystem, it's important that the encrypted filesystem is in separate files or some type of chunks, not one big blob (like truecrypt) as big blobs don't sync well when you have concurrent clients syncing.  Ideally you want a filesystem that encrypts file names, content, and inode structures separately in small efficient pieces.  While interesting, I'm seeing enough limitations and sync problems with OS X's encrypted sparse bundle approach that I don't recommend it (use EncFS if you can; else use BoxCryptor even though its closed source).

iOS and Android Support

The above approach is fully supported by iOS and Android devices using standard protocols:

• Managing email via Secure IMAP
• Sending mail via Secure SMTP
• Calendar via calDav over https
• Contacts via cardDav over https
• Network storage and sync via ownCloud iOS/Android apps; runs over over https

This probably goes without saying, but assume you'll lose your device at some point.  Think about what is on the device and how easy it is to access it.  Do you use a PIN with a self-destruct after so many incorrect entries?  Do you have logins and passwords in Contacts or Notes files?

Maintenance Notes

You will have to renew your startssl.com security certificate each year.

Spin up the occasional backup on another droplet to verify backups and the restore process works.

Security Notes

Nothing is 100% secure.  The approach I've presented here has two big problems:

• Hoards of security specialists at the big companies will collectively know more about security than you or I ever will.  Security exploits of fairly new and not widely used applications like ownCloud and davical are possible.  You're therefore effectively trading off having thousands of staff at the big SaaS providers or the government having access to your data vs relying on common sense security basics to stay safe.  In this case, we've done the basics:
• We're running the iptables firewall with only the bare minimum of ports open
• All coms over SSL
• We're not storing the actual data on the server in an encrypted format.  Ideally we'd use an encrypted filesystem on the server so that the hosting provider couldn't snoop disk data.  Of course, decrypting "on the fly" as applications access the encrypted disk is also a risk, but without using your own secured physical server you are stuck with that problem.

I've not yet installed openvpn.  Could switch access to potentially vulnerable apps like Davical's backoffice, phpmyadmin, phppgadmin to VPN only access.  I did add in .htaccess/.htpasswd files across the backoffices for slightly better security.

Lastly, this is pretty obvious, but use long passwords with lots of variation between passwords and a mix of letters (upper/lower), numbers, and symbols.

Conclusion 

Google, Apple, Dropbox and others provide a great no/low cost option for services like email, personal information management and network storage.  Signing up for an account with Google is a lot easier and cheaper than the approach outlined above.  You get most of these services "for free".  So if the thought of Google, Apple, Dropbox and others reading your emails and documents and enabling governments to do likewise doesn't bother you at all, then by all means use their free services.

However, if you think you have a right to personal information privacy without business and governments having the ability to read it then you might want to consider implementation of the approach in this tutorial.

What have I missed and what has worked well for you?

A Closer Look at BitTorrent's SyncApp

A few hours after publishing my previous blog article on p2p File Sharing - a Dropbox Killer?, I was very proactively contacted by Kos Lissounov, in charge of development for BitTorrent Sync.  I received a SyncApp tester invite and was able to test-run the product on three devices in a sync group (2x OS X, 1x Windows 7).  Kos and I also had a good and professional back-and-forth of emails and he provided thoughtful answers and comments.  It wasn't my intent to dive down the rabbit hole with SyncApp and BitTorrent security models, but some hours later...

I'm pleased to say that the "pre-alpha" version of BitTorrent SyncApp worked fine for it's main purpose at this point - quickly move around lots of files and data between a specific group of devices.  In particular, bringing a third device into the share group (all on same LAN and nicely linked together - no need to hit the cloud to download files) during my tests ran even faster with two sources of data available to bring the third device into sync.

However, to clarify the key assumption I had when writing the p2p File Sharing - a Dropbox Killer? article:  Can p2p sync replace my day-to-day use of Dropbox, but with unlimited data and at much less cost or for free?  Implicit in that assumption were all the usual points of comparison in the back of my mind: features (vs Dropbox), Just Works (stable), cheaper (for big data sets under sync), faster, more reliable, better usability, more secure, etc.

Based on this, here are a few additional comments about Bit Torrent SyncApp:

1. Empty folders are not synced.  Per Kos, this deficit is recognised and will be added.
2. Version conflicts are simply ignored.  However, rather than warning or in any way indicating a conflict, the client just silently ignores the conflicted file.  Version conflict management is apparently tough to implement.  I don't think any type of merge function is required, but I do think a visual warning and changing file names to highlight the conflict is.  Just imagine the issues trying to unpick file conflicts inside of BoxCryptor's "Package" folder with silent failure and removal of a single file from synchronisation...
3. Their is no API yet.  One could imagine an app simply accessing SyncApp folders via the filesystem on bigger devices (no API required) and using a secret to access a set of folders/files on a mobile device via an API.
4. Usability is generally a big deficit.  Features like iconic representation of sync state (as done by Dropbox) in a Finder and File Explorer window aren't present.
5. A "relay" server is required to connect devices via shared "secrets" if the devices are not on the same network and if the peers haven't otherwise previously communicated with each other.  Of course, if one device is a phone or laptop with regularly changing network parameters (on the move between networks), the relay will have to be used to link up the devices.

Items 1 and 2 above for me are showstoppers if my use case is to replace Dropbox with SyncApp.  Items 3 and 4 are painful, but might be tolerated for awhile.  Item 5's relays and "secrets" is tricky to judge because I don't fully understand the security implications.  Let's drill into relays and secrets a bit more.

Relays are a key enabler of the BitTorrent SyncApp approach.  They perform the following functions:

1. Recommend (not approve if I understand the protocol correctly) the sync of a folder/file set between devices by seeing identical encrypted secrets and recommending the two devices sync with each other.  The two devices must still directly authenticate each other (without the relay involved) - the specifics of this authentication are unknown to me.
2. Facilitate/broker communication between devices that can't otherwise discover or communicate with each other that have the same secret - deal with firewall issues just as BitTorrent clients do today.
3. Relay (SHA256 encrypted) information between two devices if the two devices can't otherwise send information between each each other.

Relays do not see any unencrypted data.  They do see and manage SHA256 encrypted "secrets".  Relays store all their information in memory, nothing is persisted to disk.  If the relay goes down then the secret relationship between devices may have to be rebuilt (depends on the device to device access and what specific functions the relay was required for in a given setup).

The concept of the shared "secret" is interesting.  It is a way to enable devices to join in a group to share content.  It has a similar feel to a Bluetooth PIN that is asserted by one device and entered by the other to allow communications between them.  Each folder (and subordinate folders and files) has a unique secret.  Relays (for now a public/shared one run by BitTorrent) are used to coordinate devices with the same secret that can't otherwise find and/or communicate with each other.

I can see two security holes with the "secret" approach:

1. A secret could be sniffed from the wire and used by a malicious SyncApp client to attempt to join a group of devices with the same secret.
2. The relay manager (for now the BitTorrent company) could manually insert malicious devices into the relay's device management system.

Kos clarified that the secret is encoded via SHA256(SHA256(secret) - therefore the password (or "secret") is stretched but not salted.  Also that it would be possible to get a list of peers but in order to join a sync group you would have to decrypt the AES256 handshake with the peer with the key SHA256(secret).  Again, I don't know the details of the protocol that engages between the two devices to actually approve joining a sync group so actual joining may be blocked.

Regardless, I remain uncomfortable with the security provided by the "secret" approach as it is today without fully understanding the protocol and implications.  This is also a showstopper for me with respect to using SyncApp to replace Dropbox at least for sensitive information.

Kos indicated a "Should device X be allowed to join this group?" a challenge will be added to SyncApp in the future to help address security concerns.  People/companies can also run their own relays meaning that they can control everything for their sync groups.  However, I think without a central service (relay or otherwise) to provide authentication and authorisation for users, devices, and secrets the product will remain limited from a usability and security perspective.  The Bit Torrent company's obsession with fully distributed, no-master, p2p approaches may really limit long-term market acceptance due to usability and security limitations.  I believe it will also limit their ability to see product adoption beyond a technical community (in which it may excel, just as it has with BitTorrent itself) and in turn be unable to monitize the product.  Even if BitTorrent did put in a central auth server (and even nominally charge for it to make money), would it be trusted given their brand position?  This is where products and companies like Cubby and Skype may have an advantage.

Even more than before, after this review of BitSync I think services like Dropbox, Box.Net, and Skydrive will struggle to compete with p2p sync as their whole business model is tied up in users consuming and paying for cloud disk space.

One use case for which BitTorrent SyncApp excels today is for a fairly technical user to simply keep a group of media files in sync, for example a photo archive on your laptop while you're on vacation being synced (automatically backed up when you have an Internet connection) to a PC running at home.    None of the above issues hold back adoption of SyncApp today for this use case.  In fact, switching to SyncApp for bulk media and other big files (e.g., install images, video, audio) and using Dropbox just for docs and simple workgroup collaboration is a good possibility for me once I'm comfortable SyncApp is sufficiently stable and Just Works.  Of course, even if SyncApp or another similar p2p product closes the feature gap with Dropbox, I still couldn't eliminate Dropbox completely because their first mover advantage is incredible and they are really bedded into the Way the Internet Works now.

If SyncApp can get past their "must be p2p only with no central auth server and keep track of nothing" view of the world and add in some of the features I've covered in this entry, I think it could become a viable Dropbox killer and be meaningful part of a "post cloud" Internet world.

p2p File Sharing - a Dropbox Killer?

I've been a big fan of Dropbox since it came out, even with the security ups and downs along the way.    They offer plenty of storage for free and it Just Works.  However, two recent announcements have got me thinking about how a competitor might go after Dropbox and other similar (mostly inferior!) products like Box.Net, Microsoft Skydrive and Google Drive.

Background - Some Recent Announcements

The first announcement was from a company called LogMeIn, and their new network storage and sync product Cubby.  Cubby's subtle twist versus Dropbox is that they offer p2p syncing called "DirectSync" (no "cloud" centralised data store required).  LogMeIn has been around for a long time and their product/service Hamachi has been a semi-tech way to set up a VPN between systems - basically a "Goto My PC" for a slightly more technical audience.  Here is the important bit: they have a long history as a p2p communications enabler (UDP hole punching, STUN, TURN, and related tricks) to make the p2p Hamachi service (and now Cubby) work in a behind NAT and firewalls.

The second announcement was by BitTorrent and their new Sync product.  BitTorrent is of course the eponymous creator of the bit torrent protocol, very successful at creating a broadly adopted way of sharing files.  It is also a p2p, no cloud required, network sync product.  BitTorrent, the underlying protocol, also has a very well established p2p communications enabler.

(Postnote: See also subsequent blog article A Closer Look at BitTorrent's SyncApp for related information.)

Now, who is the biggest, well-known p2p service out there, one with a very effective p2p communications enabler scheme, plenty of money and technical capability behind them?  Skype of course, with Microsoft's deep pockets and Skydrive capability.

Dropbox: You've been warned.

Network storage and sync: Cloud vs p2p

Of course, we don't really need a cloud service to sync directories and files between multiple devices.  rsync and boatloads of scripting around it has been been around for many years.  What Dropbox and others did was combine rsync, a cloud-based backup and control location, and It Just Works software with sufficient and simple usability to make it all work.

Skype has shown us that a p2p approach, with highly functional p2p communications enablement  to hook everything together, works just fine.  Articles on Skype discuss this very point of how they keep their infrastructure costs down by only setting up communications between two people (two devices), not actually handling the communications data itself unless they have to (setting aside group coms, super nodes and must-do coms relays to simplify a bit).  Skype wisely focused on value-added services like Skype Out to generate revenue.

BitTorrent has shown us that a flexible and lightweight p2p connection broker service can copy (sync!) massive amounts of data quickly.  And because p2p "cloud-based" components (mostly) only manage connection and data location coordination and don't manage the actual content like Dropbox does, they are light, inexpensive, and resilient (especially against content-rights litigators!).

What's the Minimum Viable Product feature set to push p2p sync over the line?

In addition to the fundamental idea of p2p sync, I think there are a few additional key features to create this new, hypothetical, revenue-generative p2p-based Dropbox killer:

1. Just Works

• Dropbox absolutely gets this one right
• With less reliance on the cloud, this should be even easier in a p2p world
• Stable (no Microsoft BSODs), fast
• Doesn't crush system performance by pigging all available disk I/O for indexing, CPU for sync calculations, Internet or local network I/O for sync transfers (if possible, sync via directly on a local LAN with no public Internet bandwidth required to shovel data between devices)
• At least the efficiency of rsync in the sync approach (e.g., block level not file level syncing)
• Works in corporate environments with strong firewalling (e.g., relay if necessary, uses typically unblocked outbound ports 80/443)

2. Pervasive - available on all high-use platforms and applications

• Desktop, Laptop - OS X, Windows, Linux
• Mobile - OS X, Android
• Application access: simply local filesystem use where possible (desktop/laptop) or an API for sandboxed environments like mobile

3. Secure

• All data sent over the network is encrypted, both to peers and to the cloud based service coordinator (e.g., public/private key)
• Service coordination:
• Authenticates you as a valid user, your account information, including billing information for billable value-add services
• Authorises your nodes you want to keep in sync
• Authenticates and authorises other's nodes you are sharing data with
• Knows nothing about your data (no relay function unless required - and even then only passes through encrypted information)
• All data is encrypted at rest in each participating node (see BoxCryptor - a big hole and opportunity in Dropbox's offering)

4. Usability

• BitTorrent really struggles here (indexing, trackers) but a viable connection coordinator (e..g, Skype, Cubby) would solve this problem.
• Tight filesystem integration (like Dropbox)
• Async recognition of folder/file changes
• Iconic representation of sync status in Finder/Explorer windows
• Clear, simple communication of sync status between devices within a client and potentially iconically:
• Which device(s) has the master/newest version of a file (think BitTorrent's Seed/Leech)?
• How synchronised is is a file, directory, or everything? (as a % complete)
• How "safe" is a file?
• # of master/complete copies (# of devices on which file fully exists)
• Are devices geographically distributed or co-located? (client asks OS for location info if available)
• Basic management of sync conflicts
• Visual notification of conflicts (Notification, Finder/Explorer iconics)
• Filesystem filename changes (as per Dropbox)
• Client logfile of existing, unreconciled conflicts with basic guidance on how to clear them
• Can flag and manage favourites and high-priority sync choices
• Camera/photo find and sync - I use this for all photo sources now

5. Sharing folders/files with others via syncing

• BitTorrent excels at this, but not in a secure way
• Use the cloud service coordinator to authenticate invited users; local node authorises access to authenticated users

What does this new p2p network storage and sync world look like?

Fast forward to a world where a company like BitTorrent Sync or LogMeIn Cubby has successfully deployed a working, free + billable services p2p sync product that has the required MVP features to compete with and beat Dropbox.

The user downloads/buys the software for each device they want to sync/copy their data between - just like Dropbox.  The client includes a simple dashboard indicating sync status and data "safety".  The client can administer all participating nodes and shared folders.  Client is a gateway to buying additional services.  User can see status of all devices, summary of content all on devices, when device was last sync'ed/seen.

Offer a user a "free" version of something that feels close to what Dropbox is today, but enables the synchronisation of an unlimited amount of data across any number of personal devices.  In fact, in general the more devices you store the data on, the faster sync happens and the "safer" the data is.

Why can't Dropbox and similar existing services do this today?

They could certainly do some of it, particularly around simplified usability.

However, Dropbox and most others have a legacy business model and investment in cloud storage, not p2p.  This will hold them back from p2p.

I'm not sure Skype can pull this off either.  Now that Skype is owned by Microsoft, one can guess the internal politics between Microsoft's Skydrive and the new Skype team will slow any progress to develop this to a crawl.  Besides, Microsoft rarely demonstrates Internet-first thinking.

What that means of course is that a new player like Cubby or BitTorrent Sync may be able to slip in.

How to make money in p2p

BitTorrent never made any money on p2p.  Just a big "thanks" from the tech-savvy Interent community for the protocol that has enabled fast and resilient file sharing for years.  However, I have a penchant for actually making money as well as technical elegance, so here are some ways that might happen with a p2p sync product.

Software sales.  Clients must be purchased for each platform you want to synchronise between.  I think users would generally accept a single shot, nominal software fee for client software purchases.  Certainly after beta and pre a mass-market ramp.  Each new device type is a new software purchase.  BoxCryptor and 1Password successfully use this model.

Services.  Much to the frustration of the media rights holders, p2p (a la BitTorrent) doesn't enable a path for them to make money.  They key of course is offering truly valuable services on top of the p2p service.

Cloud based service coordination.  I think a small fee per year could be charged after the first year for data relay, authentication and similar coordination/security services.  Certainly enough to cover related operational costs.  Looks like Cubby is doing this today.

The most obvious way to generate revenue is to create value-add services on top of cloud storage and data transfer to/from cloud storage.

A backup service is the most obvious value-add offering.

Basic backup service.  User's want to be confident that their precious documents and digital photographs are backed up to a safe and secure location.  Particularly if you build in having a separate geographic location being an important criteria to receive a "Your'e safe!" rating in the status dashboard for data redundancy.  From a security perspective, cloud based backups must be a "locked box" with only the encrypted format of the file backed up to the cloud and with only the user having the key to decrypt the files (client side de/encrypt) - not like Dropbox that store all your data in their cloud servers in an unencrypted format.

Versioning service.  Deliver as an extension to backup, perhaps like Apple's Time Capsule as a growing number of people understand the Time Machine concept.  Pay as you use for frequency and size of backups/versions.

High value local application data backup services.  Some of the below are structured data files meaning backup may not just be a simple copy.  Seek and suggest high value targets for sync and backup by offering a billable backup extension:
- Photos, iPhoto DB

- Contacts
- Calendar

- Email (local)
- Video
- 1Password DB
- BoxCryptor DB
- iTunes purchased music
- Purchased software
- ...

High value social/internet backup services.  Offer each as a billable backup extension.  A little like Facebook's timeline, but platform neutral.  Use APIs to pull out social media contributions and references:
- Email (hotmail, google, ...)
- Facebook
- Twitter
- LinkedIn
- Blogger
- ...

Web-based publication service.  Provide cloud storage and controls to Internet publish and share your content.

Specialised hardware.  Offer dedicated NAS devices that directly support the p2p sync protocol or license the protocol to NAS makers (BitTorrent Sync hits this point).  Offer from 1 to 5 disk chassis.  The sync client will automatically discover any new device connected to the local network and offer to configure it for you.  Usability will be critical.

Technical Challenges

I didn't say there aren't technical challenges here.  A few of them might be:

1. How does each node determine which file is the master among all nodes?  When to sync the newest versus flag a sync conflict and rename files.  Will have to develop a strategy around inaccurate system clocks.

2. Mobile versus desktop/laptop.  How do you manage limited space and CPU access on mobile devices versus always-connected and plenty-of-horsepower desktop/laptops?  Caching and sync prioritisation is tricky in a mixed node environment.

• User can explicitly mark high/low priority data (BitTorrent client "Transmission" has this feature) - always a high priority or just a high priority until everything in sync then back to normal priority
• User can explicitly mark favourites to imply an on-going high priority
• Automatically identify "hot" files through regular/frequent/recent use.  Files you are actively working on are implicitly prioritised for sync across all nodes.
• Amazon Kindle's approach to "cloud" vs "device" location of books is a good usability model to consider and will educate mainstream users on this model
• Always prioritise what user is asking for right now in the front of the sync work stream, ahead of what is pending (for other reasons) to be sync'ed.

Bits and pieces

There are a few other factors to consider when looking for Dropbox weaknesses.

Dropbox incurs a competitive disadvantage bourne of their very successful "share with a friend" referral model to build up membership - a whole bunch of freeloaders aren't paying for the service but still creating operational costs.  Of course, freeloaders don't cost Dropbox - the costs are borne by customers who pay for Dropbox services by paying somewhat more than the true cost of their service.  Assuming you don't sync media and just stick with documents, 5GB of free storage is room plus the refer-a-friend space bonuses can store for *a lot* of documents.

Unfortunately, Cubby appears to have copied the Dropbox business model instead of offering (e.g.) a limited duration trial and aggressive shutdown of freeloaders.  I would recommend that Cubby emphasise unlimited space free syncing for one year then charge for value-add services like data relaying and cloud backup.

Dropbox also has a higher cost by hosting their storage in AWS' S3.  They must be at a point where a dedicated in-house equivalent service with an S3 simulation wrapper around it would be cheaper.

Dropbox is still semi-technical in nature - further usability improvements are possible.

BitTorrent including their name in marketing their Sync product will be a mistake if mass-market usage is their goal.  BitTorrent file sharing has in part been inhibited from wide-spread public acceptance because of its association with illegal activity (media rights violation).  However, so long as I'm sharing my own (rightfully obtained) files… at least between my own devices… for my own use... there should be no violation.  There will have to a be a temptation with the BitTorrent approach of course to co-mingle media sharing with sync which will also inhibit mass-market acceptance.

The p2p approach to storage will take a long time to be adopted in corporate environments.  Just look at the struggle of Skype and Dropbox in the Enterprise continuing even today.

Conclusion

There is an emerging trend now to think "mobile first" in development.  Companies that are oriented toward browser based "traditional" Internet service consumption are at risk because a mobile first equivalent can come along and end-run them.  Similarly, p2p for network file storage and sync could easily become a disintermediating force for file sync and share services that currently think cloud first.  Is p2p sync a viable product in the "post cloud" world?  At least in this use case?

So who might pull this off?
- Although Skype *should* be the best horse to bet on, the Microsoft purchase, "Internet Last" thinking and internal politics may kill all hope
- I don't think BitTorrent Sync will be relevant to go mass market - too much baggage.

So that leaves LogMeIn Cubby can pull this off and steal market share from Dropbox and other "last year's tech" cloud storage and sync providers.  Of course, there are all the other MVP features above they need to get right as well, which is chancy.  Or perhaps some other startup or early/quiet competitor whose ramping up their operation right now…

(Postnote: A blog article specific to Bit Torrent's SyncApp was written a few days after this one.)

Dropbox Security, From TrueCrypt to BoxCryptor and 1Password

(If you want to skip the below and just get the recommended answer, go buy Boxcryptor and 1Password on all your platforms.  Job done.)

When Dropbox had various security issues last year (the no passwords required for some hours was the kick I needed to sort my security out), I started using Truecrypt to contain all sensitive material I was keeping in Dropbox.  Truecrypt felt good as it was opensource, free, stable, secure, and reasonably usable on OS X and MS-Win.

While I felt a 1000x better about my security situation, I also lost a lot of the convenience of Dropbox by moving to Truecrypt:

• File sync.  Truecrypt stores its filesystem in a single file.  While Dropbox is efficient at syncing big files at a block level, it doesn't cope well with changes to that file happening roughly concurrently from two or more locations.  If you mount your Truecrypt filesystem from two or more machines and make even vaguely concurrent changes (within a sync activity for example), you end up with two conflicted Truecrypt files.  One quickly learns to only open the Truecrypt volume on one machine at a time.
• Multi-platform access.  One thing Dropbox did well was to have clients available on all major platforms.  I could access my Dropbox files from OS X, MS-Win, iOS, Android and Linux.  When I switched to TrueCrypt, I was limited to PC, Linux and Mac only (and one at a time at that), no mobile/tablet access.
• Password management.  I won't say much about this other than it became harder using Truecrypt.

That was last year.  One of the great things about tech is that problems that need solving tend to get solved if you're patient enough.

Enter Boxcryptor for file security and improvements to 1Password for password management.

While there are a number of solutions available to encrypt what you store in Dropbox, I consolidated onto Boxcryptor:

• Secure.  Uses AES-256.  No cloud aspect to Boxcryptor and therefore no third party has my master key and can take a peak at my data.
• Plays nice with Dropbox.  Boxcryptor uses a folder+file structure (aka "package" on OS X) with each file encrypted separately enabling Dropbox efficiently sync.
• Multi-platform access.  Working clients on all major OSs.  At least read access on iOS and Android.
• Stable.  I've not had a single crash or corruption yet (although I'm still backing up more frequently than I might otherwise).
• No major delays in supporting the major OS upgrades.
• It allows for up to 2GB for free and more if you license it.  2GB is a lot.  Once I got comfortable with it I bought a license to get rid of the 2GB restriction.  I feel the license is a nominal cost versus the upside of more user friendly security and vendor support.

I considered Datalocker, Cloudfogger, Hyperdrive, and encrypted zip files.  All of them failed in one or more of the above.

An aside on Dropbox and sharing files:  I don't retain Dropbox's easy sharing of (encrypted) files using Boxcryptor.  Encrypted zip files still perfectly acceptable and secure way to e.g. share a single file in Dropbox with colleagues so long long as you unzip into a secure location and not into Dropbox.  Then you have to zip+encrypt and move the result back into the shared folder in Dropbox.  Zipfile usability compared to regular Dropbox sharing and syncing is poor as a result.  Note that today Boxcryptor doesn't appear to (easily) support multiple concurrently-open Boxcryptor filesystems.  When it does I could see having a Boxcryptor filesystem dedicated to sharing a set of folders/files with a specific workgroup.  Each group to have its own Boxcryptor filesystem - still somewhat painful but better than zip files.

Moving on to password management.  I have to admit my previous method wasn't overly secure and certainly TrueCrypt decreased it's usability.  As I was digging into secure storage, I also had a hunt around for how to improve password management.

Enter 1Password.  Yes, it's been around awhile, but used to be very OS X centric.  I don't know when they went multi-platform but they have.  While they've been the premium (i.e. expensive!) choice for OS X password management for awhile, the lack of support for other platforms had always been a showstopper for me.

Here is the thinking that led me to 1Password:

• Multi-platform: MS-Win, OS X, iOS, Android.  It's not on Linux, but I don't use a Linux desktop for the 1Password primary use case anyway.
• Secure.  While I can't keep 1Password's database in Boxcryptor's filesystem (I could, but I lose mobile/tablet access), the 1Password security approach is fine.  My passwords don't go to another third party password service to maintain them.  While Dropbox has my password files, they are encrypted.
• Plays nice with Dropbox.  The 1Password DB is also a folder+file (package) structure, just like Boxcryptor.  As a result, Dropbox syncing works well.
• Well supported browser plugins.  I use Chrome and Safari and both are well supported.  Support isn't quite so good on mobile/tablet platforms, but it's better than what I had before.
• Widely used.  The tech community seems to widely use it.  While not a particularly scientific measure, it seems to be on its way to being a "best practice" solution in my peer group.

I've now deployed 1Password's database into Dropbox.  It'll take me awhile to load all my credentials into 1Password but I think it's a durable investment.

One downside is that 1Password isn't overly cheap.  You have to pay for licenses for each platform (Android still free).  However, just like with Boxcryptor, I think it's worth the cost for the stability, support, and commitment to keep up with OS changes.

I did have a serious look at and play with Keepass for password management.  I like that it's free and opensource.  I liked aspects of it's design and usability.  However there were a few factors that put me off:

• Fiddly.  There are two different and somewhat competing database and application tracks, 1.x and 2.x.  Both are under active development.  There are various "unofficial" platform ports of each track to various OSs.  You have to pay attention to what version you use on e.g., OS X to make sure it's compatible with the version you use on iOS.  
• Not keeping up with OS upgrades.  The main OS X port indicated support for OS X 10.6 as most recent and today OS X is at 10.8.  I don't want to be the beta tester for new Keepass releases - what I'm securing is too critical to mess about with.
• The Keepass database is a single file, meaning that like with TrueCrypt you might have to deal with Dropbox sync collisions.

As a result, I'm an even happier Dropbox user now that I have secured files and passwords and reasonable usability to access both.  All in the licenses across all the platforms for both Boxcryptor and 1Password cost me about $125 (£80).  Yes, this is a lot, but conversely I now feel like I have the best of both worlds - the convenience of Dropbox and the comfort of strong security where it's needed.

Enabling GNUPG (PGP) with Apple OS X mail.app

(Postnote 2011-03-05: Don't waste your time on the below.  Just go directly to gpgtools mail, read the instructions, and get on with it.  It's been updated to work with OS X 10.6 and Mail 4.4.  Just tested it, works great.)

I am so not an expert on PGP, GNUPG (GNU Privacy Guard) or OS X's mail.app.  But what I can do is explain how I got the basics of PGP working with Mac mail and some resources that helped.

If you don't know anything about PGP or want more detail, see "Learn More" section at the end of this post.

The following worked for Mac OS X 10.6.3 and mail.app 4.2.

1. Install GNU's Privacy Guard (gnupg).

You need to have Macports installed.  Install it if you don't have it.

sudo port install gnupg

2. Generate your encryption key.

gpg --gen-key

Here are the options I used:

1. Option 2: DSA and Elgamal
2. Keysize: 3072 (that was the biggest keyvalue offered)
3. 0, key does not expire
4. Key identification
Real name: Jeff Blogs
email address: jeffblogs@dodgymail.com
No comment
5. Passphrase "something memorable yet complicated and long, don't share it with anyone, and don't forget it"


Your ~/.gnupg directory of configuration and databases gets set up.

3. Install the magic mail.app bundle

The bundle contains a version of GPGMail that works with OS X 10.6.3.

Exit mail.app.

mkdir ~/Library/Mail/Bundles  # if it doesn't exist already - mine didn't

Be thankful for clever, helpful and giving people and Download the bundle.

Extract from zip download and deposit GPGMail.mailbundle into ~/Library/Mail/Bundles

From the command line as the user you run mail with (not root!):

defaults write com.apple.mail EnableBundles -bool true
defaults write com.apple.mail BundleCompatibilityVersion 3

Start mail.app.

You should now have a PGP option in your mail menu (Message->PGP).

Mail.app menu with new PGP option

You should also see a PGP toolbar when you create a new email:

New PGP toolbar appears when composing a new email

(This step was the silver bullet from macrumors.com forum with an updated GPGMail from Lukas Pitschl - thank you!)

4. Create your public key.

From command line:

gpg --armor --output "Jeff Blogs.asc" --export jeffblogs@dodgymail.com

You'll need to send people your public key if you want them to send encrypted email back to you.

5. Add other people's public keys

gpg --import "Ronald McDonald.asc"

At this point you should now be able to send and receive PGP encrypted emails and mail.app will be reasonably supportive of you.

I found regularly restarting mail.app is useful when fiddling with gpg at the command line.

6. Set yourself up with a verified key service.  This will decrease warnings from mail and GNUPG.

Set yourself up with pgp.com.

Use the name and email address you used to generate your key in step 2 above.

Add the verified key service key:
gpg --import keyserver1.pgp.comGlobalDirectoryKey.asc

Let GNUPG know about the pgp.com key server.  Edit ~/.gnupg/gpg.conf and uncomment "keyserver ldap://keyserver.pgp.com" line.

(You're restarting mail.app between these steps right?)

7. Learn more!

These were helpful to the above:

• knuthbert.com - How to use GPGMail with Mac OS X 10.6 (Snow Leopard)
• sente.ch - PGP for Apple's Mail

These might have been helpful if they weren't really long, complicated, out of date, didn't work and I didn't already have the basic idea of how PGP was supposed to work:

• linuxmafia.com - An overview of GNUPG and PGP brief
• gnupg.org - The information mothership, good luck

And of course GPGMail itself, which doesn't work with current versions of Snow Leopard and mail.app.

-----

2010-06-19 Postnote: The latest OS X upgrade to Mail 4.3 disabled gpgmail.  Two things to fix this:

1. Copy GPGMail.mailbundle from "~/Library/Mail/Bundles (Disabled)" to ~/Library/Mail/Bundles

2. Enter the GPGMail.mailbundle directory and add two new UUIDs to Info.plist in the "SupportedPluginCompatibilityUUIDs" section:


E71BD599-351A-42C5-9B63-EA5C47F7CE8E
B842F7D0-4D81-4DDF-A672-129CA5B32D57


And gpgmail is working again.

(As outlined by user Bytes_U on the Apple support forums.)