Date: prev next · Thread: first prev next last
2010 Archives by date, by thread · List index


Le 2010-12-07 12:55, Michael Meeks a écrit :
Hi there,

        So - as we all know, RC1 is too large; and there are a complicated set
of reasons why that is so, which includes some tradeoff between install
time disk space, download size[1], fairness between languages, mirror
capacity, build and up-load times, and so on and so on.

        Anyhow - the good news is, that - having analyzed our .cab file by
expanding it, re-compressing each file individually (with zip), and then
analysing the results - it seems there are some fairly obvious size wins
that are possible. I've uploaded a (2Mb) spreadsheet here:

        http://users.freedesktop.org/~michael/sizes.ods

        40% of our size is endless duplicates of impress templates (one per
language) - ~none of which have any significant translated content in
them; we should simply ship one copy of each template instead. I believe
some packagers do this on Linux already.

        Similarly, ~20% of the remaining size is (once again multiply
duplicated) .ott files for the Wizards - which should be substantially
identical - with a very little translation sprinkled on top. However -
these will require coding fixes (or plain removal for 3.4 if the size is
too significant). Then another ~20% of the remaining size is license
files in English duplicated again and again per language.

        Similarly, if we cut our themes down to just two, we can save around 14
Mb on the compressed image, and so on and so on.

        Of course - this requires real work; it is not just a matter of wishing
for it :-) but it it is (I hope) all quite do-able.

        We should be able to get some of these improvements into 3.4, and more
into 3.5.

        Ergo - I am still very optimistic that we can ship a windows
installer / exe that is closer to 200Mb than 300 - without unreasonable
effort.

        Anyhow - at least for RC1 - we have (if it can be download) the code
that everyone will be running - and which is the thing that needs
testing - major regressions / crasher bugs appreciated :-)

        HTH,

                Michael.

[1] - the bigger the uncompressed .cab, the better NSIS' lzma
compression can remove duplication eg.

Thanks Michael

That does help in describing the situation. Was this also examined by other devs? I am saying this as sometimes, other pair of eyes will see other places/ways of reducing the size.

Perhaps making this exercise part of the process before releasing a major release would be a good idea?

Marc


Context


Privacy Policy | Impressum (Legal Info) | Copyright information: Unless otherwise specified, all text and images on this website are licensed under the Creative Commons Attribution-Share Alike 3.0 License. This does not include the source code of LibreOffice, which is licensed under the Mozilla Public License (MPLv2). "LibreOffice" and "The Document Foundation" are registered trademarks of their corresponding registered owners or are in actual use as trademarks in one or more countries. Their respective logos and icons are also subject to international copyright laws. Use thereof is explained in our trademark policy.