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

Re: [RFC] Switching dpkg-deb --uniform-compression by default



On Fri, Jul 15, 2016 at 04:50:27AM +0200, Cyril Brulebois wrote:
>Hi,
>
>Guillem Jover <guillem@debian.org> (2016-07-06):
>> I'd like consider switching dpkg-deb --uniform-compression by default,
>> so that both control.tar and data.tar members use the same
>> compression, which currently would be xz (or gzip with -Zgzip).
>
>(AFAICT 'none' is still supported, contrary to 'bzip2' and 'lzma').
>
>That wouldn't seem crazy to me.
>
>> This would give us more uniform and smaller packages. I think the d-i
>> people wanted something like this (?).
>
>[ Adding debian-boot@, where “the d-i people” are, and debian-cd@ for
>  completeness. ]
>
>A few years ago we pushed for xz compression in some key packages to try
>and squeeze more packages into installation images, notably CD#1; ISTR
>that would only change the data part and not the control one, which
>would limit the size gain for some specific packages. debian-cd only
>generates netinst CDs nowadays so that's no longer a hot topic for us
>AFAIK.

Almost - after some pleas we've added back a CD-only build for Xfce
too. But that's not massively relevant here I think. :-)

>> Not all .deb parsers support control.tar.xz yet, but most do:
>> <https://wiki.debian.org/Teams/Dpkg/DebSupport>
>
>udpkg's status there seems correct (supports gz/xz/no compression), and
>just to be sure: I've just checked that compression_type is indeed
>handled independently for control (in udpkg.c's dpkg_unpackcontrol) and
>for data (dpkg_dounpack).
>
>> Would there be any objections to this?
>
>Bottom-line from a d-i point of view: having both compression in sync by
>default shouldn't change anything on our side (shouldn't gain us much
>but shouldn't do much harm either).

Agreed.

-- 
Steve McIntyre, Cambridge, UK.                                steve@einval.com
"You can't barbecue lettuce!" -- Ellie Crane

Attachment: signature.asc
Description: Digital signature


Reply to: