Server Selection

Greg Whalin gwhalin at meetup.com
Thu Dec 8 14:52:31 UTC 2005


Alex Stapleton wrote:
> 
> On 8 Dec 2005, at 13:35, Greg Whalin wrote:
> 
>> Alex Stapleton wrote:
>>
>>> On 8 Dec 2005, at 10:03, Alex Stapleton wrote:
>>>
>>>>
>>>> On 6 Dec 2005, at 14:06, Gregory Block wrote:
>>>>
>>>>> On 6 Dec 2005, at 13:59, Greg Whalin wrote:
>>>>>
>>>>>>
>>>>>> This is 100% correct.  I guess I had never envisioned people   
>>>>>> putting the order in different.  I would not feel comfortable  re- 
>>>>>> sorting the list in the event that the user intentionally  set 
>>>>>> the  order for some reason (can't think of a very  compelling one 
>>>>>> off  the top of my head, but does not mean there  is not one).
>>>>>
>>>>>
>>>>>
>>>>> The knee-jerk answer:  Because that's the way it's done on  other  
>>>>> clients, and as such, behaviours between clients match?  :)
>>>>>
>>>>> I don't feel there's a good reason to sort in-client unless all   
>>>>> the clients do so; there's no point in behaving differently, as  
>>>>> it  breaks compatibility with any perl-based commandline tools  
>>>>> one  might write to access the same data.
>>>>>
>>>>> If someone wants to sort, they can sort on the way in, IMO.
>>>>
>>>>
>>>>
>>>> why not just add something like
>>>>
>>>> memcache_sort_severs()
>>>>
>>>> so that if people want to ensure they are always in the same  
>>>> order,  they can really easily?
>>>
>>>  *tries again with better spelling*
>>> What I mean, is you could just add a single method which sorts  the  
>>> server list for the user when it's called. That way they can  do 
>>> weird  ordering tricks if they really want to. It's not like it  
>>> really  *needs* to reduce flexibility. It could always be some  sort 
>>> of  toggleable option as well, so that you use the original  or 
>>> sorted  server list at will if you want to mix the two  behaviours 
>>> without  having to recreat the server list.
>>
>>
>>
>> As I see it, this would increase complexity without need.  I still  
>> can see no reason that something like this would ever be needed?
> 
> 
> Well if it's enabled by default and is set up as a toggleable flag  then 
> depending on the default (sort the list?) people who want it  sorted 
> (presumably most of them) would not have to worry about  changing any 
> code, and it would just-work (tm). Anyone crazy enough  to want to do 
> use different server list orders can always disable it  and do their 
> crazy crap.
> 
> As nobody has requested to actually be able to user different orders  
> and there isn't a particularly obvious reason to want to do it  anyway, 
> it doesn't seem unreasonable to just be lazy about it and  just 
> automatically sort it and not let people disable that behaviour  until 
> someone comes up with a compelling reason to do it differently.


Again though, this is unneeded and adds complexity where none is needed. 
  It works fine as it is now and all clients essentially do the same 
thing.  Nobody can say why this would be useful in any way. 
Adding/changing code always introduces the possibility that bugs will be 
introduced.  All I can say is that the java client will not be adding 
this feature anytime soon.  If others really want it, the client is open 
source, so have at it.  :)

Greg



More information about the memcached mailing list