[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: multiple ITPs - mbrola voices



Chris Lamb, le sam. 21 avril 2018 16:28:49 +0100, a ecrit:
> Let me play devil's advocate for a second…  Whilst I would agree
> that uploading a package containing all the existing ones for an
> addition might potentially be wasteful of mirror bandwidth, rsync &
> friends would surely be clever enough to notice the the existing
> tarballs were somewhat similar…

For the tarballs, yes, but for the resulting .debs, no. We are here
talking about 500MB worth of .debs.  Having to upload all of that just
when wanting to add a voice would mean we'd rather avoid adding voices,
which would be unfortunate.

> Moreover, human time and attention is orders of magnitude more costly
> than bandwidth

It's not only bandwidth but also disk storage.  The voices that I have
uploaded in the past have basically never changed until the repacking I
have just done, and that I believe won't do again.  If we were to have a
common packaging, it'd mean (currently) 500MB of voice data per Debian
distribution (since I guess we'd want to upload at least one new voice
per distribution, as has been the case in the past releases).

> and, for example, a packaging bug in any individual
> package is likely to be simply a bug in all them, and thus would
> require tedious and error-prone duplication of effort to address across
> 50-odd packages.

The packages are extremely simple, the fixes have always been very
easily automatized with dch, sed etc.

> And lastly, this is before we consider the aesthetic implications
> on the source package namespace.  (ie. it's ugly as hell.)

Well, that becomes a question of taste. We do need separate binary
packages anyway, I don't see how using separate source packages makes it
uglier.

Samuel


Reply to: