Tim, my fix is to *not* use the full file name.  I don't expect the 
method to check the path, only the file name.  The method is called 
checkName:... not checkPath:...

All the senders pass a lone file name in, that's what the method name 
suggests, and that (as you point out) is the only granularity at which 
you can check validity.  You can certainly write a method that tokenizes 
the path and calls this method on each part, but the whole point of this 
method is to give you back a valid filename even if the string you pass 
in contains a directory separator (as, in my example, when filing out a 
system category with a slash in it).

I agree that this issue needs (as does the rest of the file system code) 
some thought to make it perfect.  But I don't want to confuse that with 
this fix needing more thinking - all it does it put the code inside the 
method inline with its name and its senders, while making the code more 
correct and useable.

Julian

tim at sumeru.stanford.edu wrote:
> I've been taking a quick look at this and I think we need to do it a bit
> more carefully. The initial problem I see is that using the full
> filename means that it will be truncated more aggressively than now.
> Consider #isLegalFileName: for example; instead of truncating the end
> leaf of the path it would now cut the whole string. 
> 
> I suspect that the proper thing to do is to tokenise the path and do
> checkName... on each part, then reasemble the string. Or, of course,
> call the OS api to do the job :-)
> 
> tim
> 
> 
> 
> 
> 
> 
> 
> 
> 
> 
> 
> 
> 
> <This post brought to you by BFAV2>
>