[tex-live] TL 13 -- how many maintained installations are there?

[Mojca Miklavec]

On Tue, Apr 15, 2014 at 12:44 PM, AW <alexander.willand at t-online.de> wrote:
> Hi,
>
> while TL 13 will be frozen soon, I'd like to thank to all the people who
> helped creating and maintaining it.
>
> You, e.g. Karl, Norbert, and many others have been working for us, the users.
> But how many are we, the users?
>
> Clearly, you cannot count all of the tex-live installations which come with a
> Linux distribution. You never know, whether it will be used ever.
>
> So maybe we could get a rough estimation of active users by counting the
> computers which update a given TL 13 installation.
>
> Is there a way to find out how many computers have updated TL 13 using tlmgr,
> at least once? Or is this somewhere between difficult and impossible to find out,
> because the mirrors don't count how often an update has been downloaded?
>
> The introduction of the companion, 2. ed., made a forecast (I'm refering to
> the German edition): LaTeX wil be used extensivly for another 15 years until
> 2020.
>
> If possible without too much time and effort, I'd like to find out how we
> perform, ten years later now.

There's a large number of CTAN mirrors from where users can update
installation. By simply forgetting all cases when users fetch TeX Live
and update from a local server at some University, not counting any
distribution shipped by package managers etc. ... you still need to
convince admins of individual CTAN mirrors to publish summaries from
combined logs (rsync, http, ftp). And then you need to deal with users
with constantly changing IP addresses etc. Or users constantly
switching between different CTAN mirrors to prevent from counting a
single user ten times. (IP addresses are something that should not be
communicated between CTAN nodes, so I don't see any good way to
correct for that.)

You could pick a random CTAN node, do a rough count at that node and
then extrapolate. But that would be highly inaccurate metric. It might
give you the order of magnitude, but nothing more than that.

It might be interesting though to introduce an "opt-in statistics
collection package". I believe that Debian has something like that.
This could be also useful for deciding which packages are popular
among others. But it would be almost useless in the current scenario
because most users simply install all packages and be done with it.
(This might work with MikTeX, but not with TeX Live.) And if it's an
opt-in feature, you don't get any realistic numbers anyway.

So: while not entirely impossible to do it, it would probably require
some joint effort to collect at least somewhat realistic statistics.

Mojca