Destroying the incoming data did have one advantage though, if the data is deleted it can't be sent out by accident. Still, ewiki_page will return the message and destroy the data when it sees a failed auth call so I guess that's ok. > > Eeek! I have a plugin that uses the old method. When will this be in > the > > CVS? > > ;-) Already there. I changed parameters for ewiki_auth() to not use the > value backpassing, as doing so is rather inconsistent in my opinion and > would very likely break some things (the _control_links func was > affected). > > Anyhow the introduction of a general (not limited to _auth) global > variable > called $ewiki_errmsg seems a rather good thing - I wonder why I refused > that idea soo long. It is far easier to adopt (all! existing plugins need > tweaking now) than the pass-by-reference-backwards way. Because ewiki is > not OO at all, we cannot throw() anything back or so, and to keep the API > really a 'functional' one this is a better solution. > > > How about having ewiki_auth() return TRUE/'0'/FALSE on success or a > message > > on anything less than success? I prefer 0 for all it's illogic because > we > > can just test '==0', but with a definate TRUE we could use '===' to be > sure > > of the return. > > Some early -dev versions worked that way, but I feel it's far too > complicated to do so, we should go with the simple-booleans. > > mario > > > -- > broken email address: anybody who likes to consult me, could still use the > MailingList or milky@xxxxxxxxxxxxxxxxxxxxx or even 95596825@xxxxxxxxxxxxx > > > > _________________________________________________ > Scanned on 27 Oct 2003 21:12:07 > Scanning by http://erado.com