[Fontinst] Development history reconstruction

[Lars Hellström]

I'm not sure if this is anything anyone cares about anymore, but I 
anyway think I should mention it:

   I've begun reconstructing the development history of fontinst,
   in the form of a git repository.

What prompted me to do this was a request last year (I think from Karl 
Berry) about resolving a problem with TeX Live, CTAN, and teTeX all 
providing different versions of fontinst (1.926, 1.928, and 1.929 
respectively, or something like that). My first impulse was then an 
"I'd better build a release of the most recent version 1.932, then", 
but there turned out to be a catch: I couldn't find it! The directory 
where I thought I had the head of development turned out to contain 
another 1.929...  After looking in a few more places, I had still found 
nothing better, but it had become clear that searching manually would 
take time; I had _many_ versions of fontinst scattered around the hard 
drives.

Since I did encounter the tarball from which teTeX had gotten v1.929 I 
used that to fulfill the request, but resolved to sort it all out 
eventually; however I realised I would first have to do something to 
automate the process, and it was not until the last few weeks that I 
finally got around to this. In 108 directories, my script found 
instances of fontinst versions
   1.335
   1.504 (probably from CTAN:fonts/psnfss, kind of squashed)
   1.801
   1.802 (probably an 1.801 with some extra junk file confusing the
          automated scan-for-version code)
   1.901
   1.903
   1.905
   1.908
   1.910
   1.914
   1.915
   1.916
   1.921
   1.923
   1.924
   1.925
   1.926
   1.928
   1.929
   1.932 (in 6 different locations, 4 of which have different contents)
   1.933
that it then collected in a repository. The next step is to clean these 
up (since many contain unrelated files, and some are just .dtx sources) 
and reconstruct a development history from these, just for the sake of 
having one (archaeology can be fun!).

One thing that may be observed is however that there are gaps in the 
above sequence -- I'm pretty sure 1.900 and 1.904 have both been 
uploaded to CTAN, and the sources themselves speak of versions such as 
1.512 and 1.7 -- so if anyone out there is still sitting on a copy of 
such a beast, then I'd be interested to include it in the history. 
(Furthermore the 1.504's and early 1.9xx's I have are pretty mangled, 
so a proper release archive from these periods would be useful for 
reference.) I expect I'll end up lying a bit, touching up preambles 
here and there to make sure they work with available materials, and 
certainly rebuilding some directory structures, but nothing major.

The question is what to do then... I should certainly make a CTAN 
release out of the tip of trunk (once I've merged the various 1.93?'s 
into one), and I can make the repository available for download (right 
now it's about 12MB, but it'll probably shrink a bit once cleaned up), 
but perhaps a more ambitious approach would be preferable? It appears 
fontinst is still available as a project name on SourceForge...

Another issue is what to do with the non-core material that got caught 
by the initial collection. Version 1.335 in particular came with an 
impressive contrib directory, but it seems version increments >0.001 
have generally thrown away all such auxiliary materials. As there is 
probably some bitrot involved, doing so may be wise as far as CTAN is 
concerned, but in a historical archive it could be more fun to let them 
accumulate (especially if kept as a separate module). There could even 
be a point in adding other fontinst packages of current or historical 
significance, even if they have always had separate distribution: 
cyrfinst, mathfnt, and psnfss are things that come to mind. But this 
must of course be conditioned by approvals of the various authors.


Well, that's enough of me rambling! Reactions, thoughts, suggestions 
anyone?

Lars Hellström