Re: Limits on CA message size?

[winans@phebos.aps.anl.gov]

>>> Note that some clients like dm carefully keep only one connection per IOC.
>>> Others may not.  Again, Jeff and/or Andy can confirm what the demand is
>>> for each instance of a sequencer:  It uses CA but not TCP/IP, therefore,
>>> it may be smart enough to not need the buffer.
>> 
>> Why is not CA responsible for this junk?  Why should the user programs
>> have to worry about it?  
>> 
>> If CA handled this problem itself, we would have this problem solved in one 
>> single place... once and for all.
>
>Within one UNIX process CA opens only one TCP/IP virtual circuit to each
>IOC communicated with. The problem is that some applications create
>additional independent UNIX processes to manage additional process 
>variables. These applications may end creating duplicate TCP/IP
>circuits.

We know that.  But if the CA client uses the CA library to deal with the
socket operations, why can't it decide if a new one is needed or not for
each of the PV names?  Does it not already know where its sockets are
going to?

--John