[sword-devel] Beta S bug

Troy A. Griffitts sword-devel@crosswire.org
Sat, 03 Aug 2002 11:31:36 -0700

>>Not sure what you mean?  I thought [a-z,A-Z,0-9,_] was fairly explicit.
> It's explicit, but it looks entirely arbitrary.  Can you identify how you 
> decided upon this set?

Well, I can give you my justification again for this set, if you would 
like.  We need a set of values for unique identifiers that minimally 
conflicts with the least number of usages in code/markup/etc.  VCL does 
not allow control names with '~'.

The only problem we've has thus far up to this point were Christian's 
CEDICT using spaces-- which he quickly changed-- and your wordgumbo 
lexicon set that is still in beta which uses '-' and '~'.

> The changes to code that you made break existing material.  ~ is obviously 
> a reserved character for class names, but we weren't using the module id 
> to generate class names previously.

We've seen problems with these modules because the win32 frontend 
generates unique display control names based on the module name.

Now, I could filter thru the module names and change all these 
restricted characters to '_' or something else.  Then when a user 
selects a module, I could also filter the name again to match to the 
control I need to display.  And we could force all clients of the 
library to have to avoid problems such as these in a manner like this, 
or we could just fix your only module set that conflicts, and impose the 
rule for the future and save everyone alot of asperin.

I hope I'm not coming across as being difficult, but I've considered the 
above scenerios and decided it best to impose the module name restriction.

The only additional restriction I would add, which you pointed out, 
would be to limit module names from beginning with a digit (to comply 
with xmlnames.  And since you're the perl guy, there's a challenge for 
ya: come up with regex that defines the domain of a module name with the 
'no preceding numeric' restriction, so we can post it somewhere.