Hi Mathias, Christian, all, On Wed, Apr 28, 2010 at 8:13 PM, Mathias Bauer <[email protected]> wrote: > Hi Ivan, > > thanks for your suggestions. Please let me add some more technically > motivated comments. > > In the same way as developers have to care for user experience, plans > for possible changes must take technical boundary conditions into > account that at least might have an influence on when something can be > done and how long it will probably take. This can be much more than the > apparent effort that is obvious at a first glance.
Certainly. I acknowledge that things are not always as simple as one might initially believe. My case for these changes is solely on the basis of what I've been able to figure out on my own thus far, so your comments from a more technical perspective are much appreciated. > Ivan M wrote: > >> 1) Get rid of old icon sets. Industrial and Classic are two good candidates >> IMO. > > I don't know who actually uses or needs this icon sets. So not the > developers should be asked, but rather the users. I never switched icon > sets except for testing the feature, but that's only me. That is true, although I think developers should also be consulted in this in case they have any particular objection to these older icon sets being removed (e.g. maybe they are used under some special circumstances?). Users are also very important and there are 2 ways I can think of considering the users: 1) looking at the usage feedback data to see how frequently the icon sets are changed, and asking 2) consulting them - on mailing lists, the OOo forums, etc. Christian wrote that getting rid of the Industrial and Classic icon sets would save less than 10MB. That is true, but if that space saving carries over to the installer, it could reduce the installer size considerably (i.e > 5%). On further investigation with different levels of compression, I can see that there is a possibility that such space savings will have a very negligible impact. For example, if OOo uses high LZMA compression for the installer/packages then space savings will be very small because all the zip files can be further compressed from 22MB to 6MB with LZMA on high levels of compression. In that case, taking out 2 icon sets would only end up saving ~2MB in the installer. If OOo isn't using LZMA, it should :). >> 2) Run PNGOUT [2] on all icon sets to reduce the size of each PNG file. >> Small reductions across many files add up substantially. > > IMHO it doesn't make sense to save a little space in the installation > set at the price of blowing up the size of the code repository. Each new > version of an image enlarges the repository with its size, as binary > files can't be stored as a diff. Please read more about that below. Thanks for pointing that out. As I understand it, there is a branch for each release code line (according to http://wiki.services.openoffice.org/wiki/Branch_off) so that would indeed appear the opportune time to make changes. I certainly wouldn't want to inconvenience any developers. >> 3) Reduce the duplication of images in each icon set. The high contrast icons >> also appear to be duplicated in the Galaxy icon set - does anyone know why? > > I assume that not all are duplicates, so it makes sense to keep the hc > images. We won't support incomplete packages, with fallback magic that > messes up the code and makes it slower. *If* the hc images package makes > sense, I rather would expect that it should be fixed to contain hc > images only. I will look into this in more depth, but from what I saw initially, a good number of the high contrast icons were identical and even had the same file names across both icon sets. This duplication defeats the purpose of icon themes somewhat - Galaxy and High Contrast should be separate, shouldn't they? >> 4) Increase/decrease the compression of each icon set depending on the >> performance impact (i.e., we can reduce compression if it will make >> OOo load faster, or we can increase it if the extra memory use is >> inconsequential). > > See 2). Besides that I expect this effect to be unmeasurable. > >> #2 alone could save a few megabytes and would require no code >> modifications whatsoever. At the very, very least we should do this, >> and it could be done in time for OOo 3.3 (maybe even 3.2.1?) > > I have my doubts about the "few megabytes". After running PNGOUT on all icon sets, it turns out that only around 1.5-2MB can be saved because as has been mentioned, some of the icon sets have already been optimized in the past. All icon themes have some room for savings because new unoptimized images have been introduced in the meantime for each icon theme. > But even if I took that for > granted, I wouldn't believe that shaving off a few MB disk space is what > will make OOo more attractive to users. You're right: it won't be a marketable feature by any means. But if it can be done reasonably painlessly (i.e. without adversely affecting the size of the code repository and inconveniencing developers), I think it could be worth doing. The same goes for cleaning up the high contrast icon duplication. I'm happy to invest the time to do what I can - I would value the opportunity to further explore and possibly contribute to OOo. Jan has offered to help me, so I will have a better understanding of what's involved soon. Regards, Ivan. --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
