2007/587 Volo comment

[Garrett D'Amore]

Rao Shoaib wrote:
> Erik Nordmark wrote:
>>
>> I saw an issue in the list for which I have a comment. The issue is
>> djr-02    4.4.3.3 If it is possible that some protocols will or will 
>> want to
>>     support an unbind, isn't it better to provide sd_unbind() rather
>>     than rely on a magic combination of paramters to sd_bind()?
>>
>>
>> I think it is unwise to provide a sd_unbind (or a sd_disconnect for 
>> that matter). The reason is that there isn't a matching unbind or 
>> disconnect socket system call, and different protocols can actually 
>> make a different interpretation of what type of parameters to a 
>> bind() should be interpreted as semantically being an unbind operation.
>>
>> Given that there are opensource communities which develop new 
>> protocols and new mappings to socket semantics specifically for those 
>> protocols, the less semantic interpretation we make in the socket 
>> layer the better. The more interpretation we do the harder it might 
>> become to port such new protocols to Solaris.
>>
>> Thus while TCP, UDP, and RAW inet/inet6 sockets have a common 
>> interpretation of what consistutes an unbind, I don't expect other 
>> and future protocols to have the same.
>>
>> Today
> Erik thanks for your comments. The project team provided similar 
> reasoning for not adding unbind and disconnect, however after the 
> review we have added these functions. Based on Erik's comments if no 
> one objects by Spet 7th we will remove these downcalls.
>
> Rao.
>
You have my vote.  I think if you have a majority of members' in 
agreement (maybe even just a majority of the members that voted on the 
case in the first place), then you can go ahead without waiting for Sept. 7.

    -- Garrett