CAS operation

Chris Goffinet goffinet at yahoo-inc.com
Sat Sep 8 08:36:22 UTC 2007


Good call on just using the memory address as the identifier, forgot  
about that.  Where is the binary server patches svn located at? I am  
not too familiar with text protocol but I'll give it a shot. Do we  
have any C clients for the binary protocol yet?


Chris Goffinet
goffinet at yahoo-inc.com



On Sep 8, 2007, at 1:28 AM, Dustin Sallings wrote:

>
> On Sep 8, 2007, at 1:16, Chris Goffinet wrote:
>
>> Does "gets" only assign when called at that point the 64 bit unique  
>> value, or are you setting every object that is cache a unique  
>> identifier?
>
> 	I guess there's no real need to define that.  I went with 64-bits  
> because I figured we could just the memory address as the  
> identifier.  That seemed like a pretty safe bet.
>
> 	Objects could theoretically be moved by a smart allocator, but the  
> value only has value between a gets and a cas.  The worst thing that  
> could happen there is a CAS fails spuriously.
>
>> Is your prototype in python or patched diff against trunk?
>
> 	I prototyped it in my python binary server.  It's pretty good for  
> trying out this kind of thing.
>
> 	I could do the same thing in the main server pretty quickly against  
> my binary server patches, but a text protocol version might be a bit  
> tougher (not as familiar with that code).
>
> -- 
> Dustin Sallings
>
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.danga.com/pipermail/memcached/attachments/20070908/af5179dd/attachment.htm


More information about the memcached mailing list