Thoughts on https://trac.torproject.org/projects/tor/wiki/dev/SupportPolicy

[Lucky Green]

On 2011-01-25 01:09, Erinn Clark wrote:
> I think supporting multiple versions back is a laudable goal, and one we should
> consider, but right now it's not feasible (for me, anyway, as the primary
> packager with the current infrastructure). The following is my initial sketch
> of the package support policy, which is an accurate snapshot of what's
> currently out there (specifically for Tor and Vidalia):

Erinn,
My feedback is that the Tor Project really, really will want a written
and published policy of how far back an OS is supposed to be supported.
Otherwise, you will get to have this discussion every time a new OS
version is released.

Industry standard for consumer software that goes into the far corners
of the world is "current and previous major version", which has
different meanings depending on OS.

I.e, in MacOS "major version" translates to "minor version number".
Also, MacOS upgrades happen quickly in the field, which means that MacOS
can have support for older versions deprecated with less pain than most
other OS'es. MacOS upgrade uptake is to Windows upgrade uptake what
Adobe Flash upgrade cycles are to Adobe Reader's.

In Windows, "current and previous major version" translates cleanly to
major version numbers, though that means that during the current Windows
cycle you have to support three versions of Windows due to Vista's
special status of Win7 market beta. A temporary oddity and burden on Tor
Release Engineering that will soon go away.

In Debian, that translates into "stable" and "oldstable". Plus of course
whatever dev and beta versions you want to support.

Similar hardcodable rules apply to other OS'es.

Supporting older OS'es, perhaps non-intuitively, leads to reduced
overall deployment of most software as coding for less-evolved IP stacks
and GUIs takes away engineering resources, constraining beneficial
features that would attract more users.

--Lucky, wearing his day job hat for brief moment.