Two binary protocol questions (draft, stats)

dormando dormando at
Thu Apr 3 21:43:53 UTC 2008

Dustin Sallings wrote:
> On Apr 3, 2008, at 11:51, dormando wrote:
>> Some basics, like the remaining bytes disagree with the draft. 
>> Obviously it's less work to just update the draft, but are we missing 
>> features/intentions? I haven't had time to look, hoping Dustin knows :)
> I'f flagged the patches and comments as things to look at, but I haven't 
> actually done so yet.  I'm hoping for ``soon.''
> At what point (past or future) do we want to consider the protocol in 
> the server ``done?''  That is, if something is inconsistent and ugly in 
> a non-major way (e.g. I'd really like to not do a major rewrite of the 
> headers *again*), it's probably good to get it fixed before everyone 
> switches over (which I'm sure is next week, so we should act fast).

I listed most of my own (ideas, I guess) on that along with the 1.3.0 
release announcement. With flagrant musts and must'n'ts.

Wouldn't kill us to at least review what parts of the docs are 
inconsistent with the server, laying them out on the list, or just 
heavyhandedly deciding they differences are meaningless and updating the 

I guess the bigger deal is to make stats work. Trond's apparently 
already fixed the syscall performance issues, although we have not 
verified if there are raw CPU regressions or the like (also detailed in 
my release notes).

But yeah, the 1.3.0 series is open for minor changes. You don't have to 
do them yourself if you're busy though, since it was me who put it out 
as such :)


More information about the memcached mailing list