A site devoted to discussing techniques that promote quality and ethical practices in software development.

Tuesday, December 15, 2015

Is building a better mouse trap (Signal Private Messenger) enough to win market shares?

I am please to see the release of Signal Private Messenger for Android and iOS, a messaging application that has earned full marks in the EFF security score sheet. I am a fan of this product and I like it very much for the following reasons:
  • It is an open-source project offering the service for free. WhatsApp is not a free.
  • As a result, it can be reviewed by anyone capable of doing it while WhatsApp is proprietary, even though it claims to be underpinning by Open Whisper Systems but no one has reviewed that. Recent event has indicated that WhatsApp messages have been intercepted and decoded.
  • It is not owned by any company while WhatsApp is owned by Facebook, Skype by Microsoft. Thus all metadata in WhatsApp and Skype belongs to Facebook or Microsoft respectively.

According to well-known security researchers, Bruce Schneier and Matt Green, Signal is developed to a very high quality to provide end-to-end encryption (E2E) not only for messaging but also for voice and their endorsement must mean something.

I am not here to raise doubt of this product which I am using admittedly with very limited users to interact with and I have great trust. I hope it will do well.

But I am here to question whether it is enough to rely on technical superiority which is so well hidden from the users to induce them to switch to Signal and to grow its market shares. That's is: is building a smarter (more secure) mouse trap enough to win market shares? Other class of software such as web browser, anti-virus, media player, or mail client can draw people to switch based of superiority of features.

Looking at the landscape of messaging applications it is difficult to see how Signal can rely on security implementation, so out of sight of the user, to win market shares. Will this become a replay of VHS (WhatsApp, Skype, etc) vs BetaMax (Signal) of the 21st Century?

Messaging applications are like clubs or cults in which they only allow club members to interact and go to great length to discourage inducement to leave and definitely providing no facility to support inter-club interaction. This produces network effect to draw people in and that also becomes disincentive to leave and its nurture of human social interaction provides a positive feedback to increase the network effect.

Looking at the EFF Security score card, most of the popular messaging applications do not use security best practices and their inferiorities do not seem to matter to the users. The anecdotal conclusion one can draw is that users do not care with online privacy and security despite well publicised massive surveillance activities. Unlike other type of application, such as web browser, there is no report of people deserting one messaging application to another, despite vulnerabilities and caught not using secure messaging mechanism when they claim to use. For those entrenched players, they must feel like in a no-loss situation. The only way they can lose to a competitor is by a total annihilation of the enterprise.

Messaging applications have another unique characteristics that it is not the features that draw users to choose a particular application; there is a great degree of peer pressure exerted by those early adapters unwittingly forcing people to form that circle of friends. This peer pressure then forms a vortex to draw more and more people in. Their only concern is to be able to communicate with the club members.

Because of the lack support for inter-application interaction, the application through using proprietary communication protocol forms a natural barrier for their user to leave. Apart from that, the user does not see any benefit for using a different application that essentially providing the same things - messaging and may be voice - and having to desert their friends. So why leave? What is the benefit to them?

Many users of messaging applications also form the mistaken belief that they can only use one messaging application in their device. Perhaps it is this mistaken belief or blind fanaticism to their favourite application they are also reluctant to install other messaging applications to increase their reach to their friends. Since Signal is so similar to WhatsApp, it is simply a matter of installing and waiting for others in the contact to install their copy of Signal to re-establish communication. Even that simple is not enticing.

I have spoken to several users of messaging applications as well as non-users and recommending to switch over to a more secure application called Signal. But telling them the benefits of Signal is like talking about wine apprecThis is particularly difficult when Signal is so similar to the operations of WhatsApp separated by a thin veneer of technical features. In view of this, users of WhatsApp (or other app) are unwilling to desert their circle of friends to use something that to them is almost the same thing with minute user base, by comparison. iation to a group of teetotalers. To them the improve security and end-to-end encryption (E2E) are not enough to sway them. Even people that has not used messaging application seems to be reluctant to get onboard with Signal because they have not heard of it being mentioned by their friends.

So I wonder how a late comer like Signal can overcome these barriers to increase its market shares? How it can base on technical superiority to entice users who are disinterested of them that Signal relies on to distinguish it from others? What is the future of Signal apart from being a niche player at best? Clearly Signal needs to improve its image and marketing.

From the analysis, users of messaging applications place extremely high premium on their ability to reach their circle of friends and ignore other issues like security and privacy. Therefore if the new comer, like Signal, wanting to rise up, it must give their users a transparent way to interact with their circle of friends without requiring them to switch en masse like the present situation. How to achieve that is the real challenge in messaging application development in view of no standard communication protocol?

Monday, December 7, 2015

Comments on using e-mail address as username for online services

I have encountered more and more online facilities using e-mail address as the user name. In my mind, this is a lazy way for the service to check or to provide a unique user name when creating an account. In some rare usage, this may be fine but generally this is a very restrictive form and the reasons are given below.

Using e-mail address has the following problems:
1) While it is unique in the universe of the Internet it does not uniquely identify a user of the service, thus unsuitable as a user name unless the service has other facility to deal with one e-mail address for multiple users.

For example if one manages several properties or funds belonging to different entities under some management agreement, it is often convenient to use one e-mail address for all these properties or funds. It is also possible that the e-mail owner owns all those properties or funds, it is unreasonable to base that identifier on an e-mail address which does not map to a unique entity; e-mail address is for correspondence - like a house address.

Who would then use a house address to identify a person living there when it can house several persons?

I have seen one service that uses the user name (aka e-mail address) as a proxy to a fund account. This then assume the owner of that e-mail address cannot have more than one funds - one may be for him and another for some other ownership arrangement with correspondence being sent to the same address. Clearly the developers have not model the usage requirement well.

This silly design is like the above house number analogy requiring a house to house just one person.

The assumption that an e-mail address uniquely maps to a particular person or entity is unsound. Don't do it. It is far better and more secure if your system generates a unique number, a la, account number, for the user.

2) The use of e-mail address as a user name can confuse user in that he/she has to supply to the online service the same password for e-mail account. This can lead to an increase (or subliminally encouraging) reuse of password, a dangerous practice.

To a less technically savvy person, he/she may be misled into believing that the e-mail provider now have access or linked to whatever materials available in the online service.

3) While it is infrequent, though not impossible or improbable, for people to change e-mail address, services that uses e-mail address for correspondence as well as for user identification inevitably prevent user from changing e-mail address. This is because it is using a very poor design pattern - one piece of data to serve two distinctly different and diverse purposes. The user name is to identify a user which an e-mail address does not and the e-mail address is for correspondence, like a house address which can be used by anyone living there to receive correspondence.

If you ask correspondence sender to simply put the address on the envelope no one in the household will know to whom is that letter addressed; you need to put the addressee's name (the user name). A person could one day moves out of that address; he/she retains the same name (user name) but simply changing the delivery address (changing the e-mail address). This happening may not be frequent but not improbable or impossible.

No right minded person would combine the two (addressee's name and the delivery address) but why do that in the computer system?

To address this kind of short coming, they then have to provide a means for the user to define an e-mail address for correspondence. In this situation which one should the system uses during account set up and validation purpose?

How to overcome this poor design as a user?

If you, as a user, are confronted with this problem - how to use one e-mail address for more than one users of the service - you may try this solution provided that:
  • Your e-mail provider supports e-mail alias. GMail and Hotmail support them. If you provider does not supports this, set up a GMail account as a mail redirector.
  • Your online service's user name (aka e-mail address) validation knows about RFC 822 - Section 6 Address specification. Those failing to parse this properly would reject your e-mail address with alias.
Then use e-mail alias (like Somebody+Property1@GMail.com or Somebody+Property1@Hotmail.com) to allow one e-mail address to be used for several entities. The '+' character in the local part of the e-mail address is valid and permitted under the RFC. If their developers tell you that it is an incorrect address, point them to the RFC.

Those thinking of using e-mail address as a user name to relieve them the task to validate its uniqueness needs to validate the e-mail address to conform to the RFC.

To me, the task of validating and ensuring a user name is unique within the system is far easier than validating the e-mail address because the latter needs to check:
  • conformance to RFC
  • that the e-mail provider supports the e-mail alias that the user enters, as the service has to make sure it is a reachable address to receive correspondence. If that alias syntax is not supported by the mail provider, conforming to RFC does not guarantee it can be used for correspondence.  
Here lies the danger of tying the two purposes to one piece of data, that is using an inappropriate design pattern.

Blog Archive