[Libreoffice-qa] minutes of ESC call ...

Pedro pedlino at gmail.com
Tue Jan 8 16:54:59 PST 2013

Hi Michael

Thank you for the season greetings. I hope yours were good as well ;)

Michael Meeks-2 wrote
> 	That's the current plan. The original decision was that creating 100x
> the size of what we have now creates significant up-load and build
> performance problems: currently we can sync. to our mirror network
> reasonably quickly. Failing that we'd have to try to pick some subset of
> languages: which seems likely to annoy translators a lot - and bust our
> attempts to "treat all languages equally" :-)
> 	Taking just the Win32 build - 140Mb * 100 is 14Gb of data to up-load
> and sync. between mirrors: and we do more than Win32 of course :-)

I don't understand the emphasis on "quickly". "The other" free office used
to have some 70 languages available and they didn't seem worried about
BTW 14Gb is peanuts for an organization such as TDF. Individual users (with
home grade internet connections) are transferring several Gb movies and
games by bittorrent every second...

Michael Meeks-2 wrote
> 	The download for 4.0 is (AFAICS) 184Mb - which is some 21Mb smaller
> than the previous release - (for more features :-). So - I'm not sure
> that the download time difference is an issue. More importantly - since
> it can be distributed to mirrors fast I suspect we have an overall
> better download performance in many non-dial-up cases ;-)

Yes, but 184Mb is 50Mb (27%) larger than some 135Mb for a localized build
and still it DOESN'T include the Help package which is another 10-20Mb.
Having two separate downloads is awkward (especially because not all users
get it from LO.org where it is more obvious that you should get both)
This is not the first time I answered this question and I fear it won't be
the last

Michael Meeks-2 wrote
> 	Of course - we still have a lot of fat in there that can be trimmed: a
> lot of that got cut out for 4.0 but I expect there is more: if download
> size is the issue - then more profiling and work to improve how we do
> our translations is prolly the most effective way to help here.

Thirty percent differences in the download size in the current update model
is probably the least important factor. The problem is that the package is
not complete and that you have to get the whole thing every time. Of course
it is essential that LO moves on to incremental updates (at least under
Windows) as "the other" free office is moving to. 

Michael Meeks-2 wrote
> 	That's a shame indeed. I wonder if there is some good way to address
> that problem. Could we add some "install new languages" setup option ?
> presumably this is configurable on the command-line and in the setup UI
> (?)

If TDF insists on multi-language installers, then the installer's first
screen (I have suggested this a million times...) should be a Language
selection box (with a suggestion based on detected location) but which still
allows the user to decide (instead of taking the decision for him in
exchange for saving a couple of mouse clicks). This does not require
reinventing the wheel. Probably the installer used by LO already has this
option. See the Inkscape installer as an example.

Maybe create some 10 separate installers which include the localized program
and help files? Say the first covers languages from ast to cs, the second
from da to et, etc...

Still the best option is a small single-language installer with help files
included. All other options IMO are workarounds.

Michael Meeks-2 wrote
>> (BTW out of curiosity it is interesting to see the number of downloads
>> per
>> OS
>> http://sourceforge.net/projects/openofficeorg.mirror/files/stable/3.4.1/)
> 	Not sure how to interpret that :-) We have per-language download
> statistics (of varying reliability) if you're interested in that: there
> is a clear long-tail effect for minority languages; or was it
> per-platform you want ? [the vast majority are Windows of course].

No second intentions. I'm just curious. Is there some place with similar RAW

Michael Meeks-2 wrote
> 	So - in summary, there is significant cost to this; it may be the right
> thing to move to per-language downloads: but if so given the significant
> cost of producing them: we'd need to better quantify the benefits of
> that: how much of our download size is languages currently ? how many
> people need the off-line help that don't have it (that's rather harder
> to know), and so on. Having someone working on that would be great.

If there is a log of downloads (even if IPs are encoded to preserve privacy)
it should be possible to check if the same PC got both packages (as long as
it's on the same hour/day)
I'm sure someone would be happy to do some data mining :)


View this message in context: http://nabble.documentfoundation.org/Libreoffice-qa-minutes-of-ESC-call-tp4024205p4027876.html
Sent from the QA mailing list archive at Nabble.com.

More information about the Libreoffice-qa mailing list