Re: [Pan-users] Launch feedback, splash screen?

[Charles Kerr]

On Sun, Jan 12, 2003 at 04:22:51PM -0700, Duncan wrote:
> I really enjoyed the faster start times on some of the versions a bit ago, 
> with group loading in the background.  However, I did see the confusion it 
> caused when a user loaded a group that wasn't fully loaded by PAN yet, and 
> there were mentions of even worse race conditions, so reverting to the 
> preloading was probably for the best.
> 
> However, that doesn't mean it's good when PAN takes so long to start with no 
> indication of what's happened, for those of us with large caches.  Some sort 
> of a splash screen might be nice, to give some indication of what's going on. 
>  
> Just showing the PAN about box (which could possibly be dressed up with a 
> mascot image or something,  image of PAN?) with an additional "Loading 
> messages..." line would work.  Since some folks don't like splash screens, it 
> should be a toggleable option.  Ideally, it would have a progress bar across 
> the bottom, so the user could watch the load progress, but that's more chrome 
> than absolutely necessary.

This is unnecessary as of 0.13.3 -- Pan now initializes the cache so fast
that splash pages and cache worker threads should be  unnecessary.

Here's the boring, technical detail:

What the cache needs to know on startup is this: "what are the message-ids
of articles cached locally?"  Originally Pan just used the message-id as the
filename, on startup it just walked through the cache directory collecting
filenames.  Many users pointed out that some articles often message-ids
containing characters invalid on some filesystems.

The second approach two was to strip out those forbidden characters and use
the stripped form as the filename.  On startup Pan walked through the directory,
opened each file, parsed its headers for a Message-ID, and closed the file.
This made startup very slow when the cache was large.

0.13.3 goes back to the first, fastest, approach by changing the transform
from a lossy one (stripping out all forbidden characters) to a non-lossy one
(encoding the forbidden characters), like the way browsers convert spaces in
URLs to %20 . So it doesn't have to open, read, parse, and close
the cache files anymore.

cheers,
Charles