Re: Making balsa suitable for small devices



On 08.06.2004 11:20:45, Wookey wrote:
Hi people, I'm an arm-linux, Debian and Embedded Debian developer and
I'm
involved in a project that wants to use balsa on a small web-pad type
device
(but with a real keyboard).

Our biggest constraint is flash size - we've got an awful lot of stuff
to
shoehorn into ~60Mb. Balsa seems to have a pretty reasonsble footprint
for
it's functionality (better than sylpheed which is the main competition
:-)
), however it's still too fat.


What I'm hoping I can get from you guys is some clues on what options
there
are for minimising the footprint, and help with making it good to use
on a
touchscreen (only one sort of click, no tooltips, big icons).

Presumably there is room for some shinkage by missing out libraries or
functionality or languages? What tradeoffs are there are to be made
(both
easily or with some development). Does it have a --without-gnome
option
(like gnumeric) to remove a load of gconf and bonobo stuff for
example?


The things we definately need to keep are HTML mail (spit - I hate
it!) and
POP and IMAP, and decent fonts. Much of the rest is up for debate.

There will definately need to be UI changes for the touchscreen use -
presumably you'd be happy for these to come back into the mainstream
for
others that need this? We're keen for our changes to go upstream
wherever
possible as it makes our lives easier, but there may be some things we
want
to do which aren't really what the balsa team is looking for. I need
guidance on that.


One problem we have right now is that it's a bit flaky - it crashes
far too
opften (we're testing with the Debian build 2.0.17-2) and it's running
on an
arm-based device. I don't know how much testing has been done on arm.
Is
flakiness expected?


And finally there is a possiblity of some paid work making said
changes.
Anyone who is in a position to work on it _this month_ should tell me
what
they can do, when they can do it, and how much cash they'd need to
concentrate their minds (off list is probably best). You would have to
sign
an NDA to keep you quiet about the details for the next 3 months or so
(These corporate clients don't like anyone to know what they are doing
:-)


No promises at this stage, but we have a lot to do in s short time and
using
the relevant experts whilst giving something back to the community is
the
best way in our experience.


Sorry to come barging in here in such a businesslike fashion - I hope
I can
contribute something useful to the project one way or another.

One more comments wrt to crashes for 2.0.x: you should compile without threads, that would make things much more stable.
./autogen.sh --disable-threads
Bye
Manu




[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]