bug in PHP API, with fix attached

Justin jmat@shutdown.net
Thu, 4 Dec 2003 23:40:56 -0500

The new one (http://phpca.cytherianage.net/memcached/) is much faster than
the original one.  The newer one uses persistent sockets and streaming --
the first/old client was just a direct port of the original Perl client (by
Ryan Gilfether).  As of August or so, he abandoned development on it... 

Ryan Gilfether offered to let me take it over, but Ryan Dean beat me to it
by releasing a (far superior) client.  Ryan Dean's client is about 70%
faster on multiple read/writes because of the persistent sockets and

Brad, if you can contact Ryan G (his email bounced the last time I mailed
him), I'm pretty sure he'd be more than okay with the official abandoning of
his client. Also, it doesn't work under PHP 5 beta 2 because of the major
class/object changes, and I'm assuming it won't work properly under PHP 5
production (ETA January). Ryan Dean's client works flawlessly (and is even
faster under PHP5 than PHP4).


-----Original Message-----
From: memcached-admin@lists.danga.com
[mailto:memcached-admin@lists.danga.com] On Behalf Of Tim Starling
Sent: Wednesday, December 03, 2003 6:49 PM
To: memcached@lists.danga.com
Subject: bug in PHP API, with fix attached 

Brad Fitzpatrick wrote:

> Tim,
> There are two PHP APIs.  It's my understanding that this one is
> streams-based and doesn't have the bug you mentioned:
>    http://phpca.cytherianage.net/memcached/

Well now neither of them have the bug I mentioned. At least in our copy 
:) Brion Vibber told me there was another client, but I chose to ignore 
him and fix the old one instead. It's more satisfying.

> That said, I'll wait for the maintainer of the original PHP client to
> accept this patch or let me know if alternate PHP client should be the
> de-facto one.

You want the maintainer of the original client to tell you whether the 
alternate client is better? I have a feeling said maintainer may be 
somewhat biased in this respect.

In any case, the bug I just fixed in the client was really pretty minor 
compared to the lack of automatic slabs reassignment. That's a fix which 
is sorely needed.

BTW, can you please configure mailman so that it obscures email 
addresses in the web archive? Also, sorry for the duplicate post.

-- Tim Starling