On Tue, 20 Jan 2009 14:43:12 +0100 Loïc Minier <lool@dooz.org> wrote: > On Tue, Jan 20, 2009, Дмитрий Ледков wrote: > > My understanding was that checksums are provided for > > original tarball, diff, and deb. Not the individual files in the > > package, hence my logic was that no checksums will be lost. > > I meant the /var/lib/dpkg/info/*.md5sums to use with e.g. debsums. Is there any actual evidence of a problem with the current situation? Changing the endianness of a .mo normally requires the original .po and the msgfmt binary which means a runtime dependency on gettext *AND* the need for the original source. Even with another method of conversion, what is the possible gain and why is this only a concern for one package when every gettext translation is theoretically involved? In short, why bother with the entire idea? It's only useful on very limited devices, at which point you should be looking at Emdebian. As far as web browsers are concerned, it would be better putting your efforts into getting dillo2 into shape for Debian so that users who care about this tiny performance issue are not wasting resources on epiphany in the first place. (dillo2 is the Gtk2.0 version of dillo but uses a statically linked unreleased toolkit library currently). Any device where the user cares about the endianness of a series of .mo files is going to get a lot more benefit from having a smaller web browser in the first place. If epiphany is really concerned about performance on low-resource machines, maybe more effort on reducing the dependencies would be the solution? Flipping the endianness of a few .mo files at startup isn't going to have a significant impact on startup time on an average Debian machine. -- Neil Williams ============= http://www.data-freedom.org/ http://www.nosoftwarepatents.com/ http://www.linux.codehelp.co.uk/
Attachment:
pgpBxkw8xfOko.pgp
Description: PGP signature