Binary protocol document as of 3:05AM at Yahoo's hackathon

Dustin Sallings dustin at
Sat Dec 15 20:13:21 UTC 2007

   It's a semantic difference.  We want to have as few semantic  
differences as possible.  We *did* add a couple if items that were  
available elsewhere in the text protocol in ways that were redundant,  
but they don't add functionality.

   In the end, we'd like to avoid telling people they get some feature  
with one protocol, but not the other.

   The right thing to do here might be to add a getattr call and a  
corresponding "touch" command to update flag and expiration.

   I'm still not convinced knowing the expiration time is all that  
useful in general.  Data may still be overwritten, deleted, or  
modified independently of the TTL, and it's not guaranteed to last  
that long.  Without positive confirmation of invalidation cmfeom the  
server's point of view, you'll end up with all kinds of cache  
consistency problems.

Dustin Sallings (mobile)

On Dec 15, 2007, at 8:01, Nicolás Lichtmaier <nick at>  

>> Attached is the most current documentation of the binary protocol  
>> after
>> hammering through more details at the Yahoo! hosted hackathon today.
> Wouldn't it at this stage be very easy to add 32 bits to the  
> response to send the "remaining TTL"? Any drawbacks to that? (Sorry  
> for being repetitive =) )

More information about the memcached mailing list