Re: LSR

[Graham Percival]

On 7-Jul-05, at 10:47 AM, Erik Sandberg wrote:

> On Thursday 07 July 2005 12.14, Mats Bengtsson wrote:
> > In contrast to non-trick-worthy? For example, I find most of the
> > examples in the current Regression Test very useful as Tips and Tricks
> > as well.

> If we would integrate all of LSR into the manual, it would also mean 
> that we
> would have to maintain all those snippets. Therefore, it might be good 
> to
> moderate the collection somehow.

As far as maintaining them... I'm not too concerned.  AFAIK we don't
really maintain the input/test/ stuff.  If somebody points out that 
something
isn't working, we fix it, but we don't go looking for mistakes.  Having 
a lot
of user traffic on LSR would be a step up.

I like the moderation, though; I could imagine using four or five 
different
levels:
1- example came from the manual.  We don't include it anywhere in our
generated docs, but they're still in the snippet tarball, and could be
included in grepping.
2- example came from input/regression.  This is a bit tricky, since I 
don't
think that the programmers want to deal with updating LSR whenever
they want to add a test case.  My instinct is to leave input/regression/
alone.  If LSR wants to include them, fine, but I think that 
input/regression/
and LSR-level 2 stuff would get out of sync.
3- easy tricks  (mostly from input/test/, but new stuff as well)
4- advanced tricks
%%  we either guess at "easy" and "hard", or split it into "non-scheme" 
and
%%  "scheme" tricks, or something.  Currently input/test/ isn't too 
big, but
%%  since LSR is going to be a huge success, we're going to get tons of
%%  extra snippets, so we should plan for future expansion.  :)
5-  long tricks

Cheers,
- Graham