Re: [Libreoffice] update service

Hello Arnaud,

On 01/05/11 15:26, Arnaud Versini wrote:
> I've tried this system with some problems :
>
>     * I can't build it, it needs some notable modifications before
>       being usable with wxWidgets 2.8.
>     * this solution needs wxWidgets, I dont know if it's a problem but
>       LibO download size increase with.
>
> Is it the good way?

Thanks for looking into it. Some points:

1) I'm not at all sure WinSparkle is the way to go, given that it only 
works on Windows. Even if the work is relevant to Mac as well, given 
that WS is 'heavily inspired' by Sparkle, if it doesn't do at least *nix 
as well we're going to end up with parallel update systems. I think we 
should continue looking for a cross-platform solution for this.

2) OTOH, I haven't looked into how Sparkle / WinSparkle work, 
particularly how they package updates. The appcasts concept sounds 
simple and elegant, which makes it smell technically attractive to me as 
well, so my only question mark is how updates as actually packaged / 
applied. IF, through study of the Sparkle / WinSparkle systems, it turns 
out that the same packaging / patch applying method could work on *nix 
as well, then maybe this could be a fruitful direction to go in.

3) From my extremely limited knowledge of Wx, I know that it's far from 
monolithic. There are umpteen shared libraries that all do various bits 
and pieces. The problem Wx might pose would depend on what bits we need. 
I know there are 'low-level' Wx libraries that abstract system 
utilities-type stuff, which can be used entirely separately from any 
graphical libs. Depending on what WinSparkle needs one might be (should 
be?) able to do the GUI part using GUI components already in LO.

4) There have been proposals of refactoring all GUI components out of LO 
and using a third party GUI library. Some people recommended Qt, for a 
variety of reasons I'd recommend Wx instead. It would be an ungodly huge 
job of course, but maybe the experience of tinkering with incorporating 
an external GUI lib with LO code would be profitable for future development.

5) The other way to go is just examine how the (Win)Sparkle system works 
and emulate it using LO components. It'd be cross-platform because LO 
components are already. Again, the magic would seem to be in how updates 
are packaged and applied.

... make of that what you will :-p

-r