Re: Gnome Journal Article and General Beagle Publicity

[Joe Shaw <joeshaw novell com>]

Hi,

On Sun, 2006-11-12 at 20:59 -0500, D Bera wrote:
> Only if you dont do the rounds of distro bugzilla and newbie forums.
> Those places are full of beagle eating CPU and memory problems (some
> of them are due to some old buggy version of beagle that hasnt been
> updated in the distros). I dont think beagle is newbie ready i.e.
> plug-and-play yet.

I think we're addressing these issues pretty well.  We've been shipping
Beagle on by default on SLED 10 and openSUSE since 10.0 and the reviews
have been largely positive.  There are still issues with certain file
types, but I truly feel that Beagle is quite usable and useful by
default.

> Kevin, I had a brief glance at the thunderbird backend. I didnt see
> any obvious place which takes a lot of memory. If the concern is the
> ReadToEnd() in the mork parser, then you can try to replace the huge
> content string with a StreamAsArray type implementation. I replaced
> one ReadToEnd() in the html filter in a similar way, check
> Filters/HtmlAgilityPack/HtmlDocument.cs - it contains the
> implementation. Might be useful.

The ReadToEnd() is one killer in the Thunderbird backend.  I have a .msf
file that is 11 megs; so reading it in all at once more than doubles my
heap size, which isn't acceptable.

The hash table Kevin talks about is also a big problem, but I don't know
the specifics behind it.

By the way, there is a new Mono heap profiler in Mono SVN called
heap-shot.  Unlike heap-buddy which is a summarizing profiler, heap-shot
lets you take snapshots of the heap at different times and lets you
compare them.  This lets you know what types are using all of your
memory at a given time and lets you track the changes over time.

Joe