If an exact match cannot be determined, should an exception be thrown 
(fail-fast?), or should some other behavior be implemented -- such as 
selecting the default sub-application?  (I keep wanting to say 
defaulting to ... and that just gets to be a tongue-twister!)

I'm going to implement this.  I view it (as do many others) as a bug - 
think someone even filed on it.  I'd love some "expert direction" on 
this certain behavior though.  Oh - you'll get a diff -u once it's done 
:-)  Of that you can rest assured.

Thanks!

Eddie Bush wrote:

> Why do we even need to lookup the prefix?  That seems error-prone, at 
> best.  Yeah, ok, maybe we should check to be sure that what we arrive 
> at for a prefix is valid, and ... throw an exception if it isn't.  I 
> think we should select the application differently.  What I would like 
> to do is:
>
>    - get the index of "/" (beyond the first one)
>        - if it's not there, this is the default sub-application
>        - if it's there, get the substring from the initial "/" - up 
> to, but excluding, the final "/"
>            - lookup this substring in the prefix array (exact match)
>                - if it is present, we choose that sub-application
>                - if it is not present, we assume the default 
> sub-application or throw an exception
>
> This way, we always default to the ... default ... sub-application.  
> So, if someone had a "/" in the querystring (for whatever reason), we 
> would fail to find a match with our declared prefixes.  The default 
> sub-application would be selected.  Say we manage to get a match - 
> that sub-application would be selected.  Also, we have just (greatly) 
> increased the namespace for possible sub-applications - compared to 
> what it is right now.
>
> How does that sound?  If I should throw an exception, rather than 
> default to the ... I don't like saying default for the sub-app name 
> ... rather than selecting the default sub-app, what would it be?  An 
> unavailable exception wouldn't be appropriate at this point, I don't 
> think.  Am I wrong?  I'm open to suggestions :-)  Does this sound like 
> something you'd accept a patch for? 


-- 
Eddie Bush




--
To unsubscribe, e-mail:   <mailto:[EMAIL PROTECTED]>
For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>

Reply via email to