[Click] ToSocket question

[Eddie Kohler]

Hi Mark, Yun,

I'd love to see your current patches to To/FromSocket!  In particular, 
Yun, what'd you do to UNIX domain sockets?

Eddie



On Mar 7, 2005, at 2:29 PM, Mark Huang wrote:

>> I'm trying to understand why ToSocket is an agnostic type. It seems 
>> to me that it should behave pretty much like ToDevice.u One big 
>> difference is that (from my understanding from the source) ToSocket 
>> will keep pulling packets from the input, no matter whether the 
>> socket is available to send stuff or not. I found this could block 
>> click in sendto(). And because it will pull all packets from the 
>> input, Queueing won't help.
>
> I wrote ToSocket; originally, it was a direct derivative (read: stolen 
> goods) of ToDevice, which was agnostic, but I fell behind in 
> maintaining ToSocket and it looks like ToDevice is now l/h. The 
> behavior your mention should probably be the default, i.e. open a 
> non-blocking socket and pull packets when the socket becomes 
> available. Patches welcome, I've been meaning to submit some recent 
> changes I made to it, to Eddie.
>
>> Another question: Did anyone try the UNIX socket before? I had to 
>> hack the sendto call to send because my Linux complained that the 
>> socket is connected and won't do the job.
>
> I tested it at one point when I was writing it, but haven't used it in 
> a long time. Again, patches welcome. I'm a little busy right now, but 
> should have some time for Click very soon...
>
>> BTW, First week in click, I have to say cool stuff and fun!
>
> If you're using Click on PlanetLab, let me know. All of the elements 
> I've contributed are directly related to PlanetLab. I'd like to start 
> a topic on the PlanetLab Wiki to document some of the interesting work 
> that's going on with Click.
>
> --Mark
> _______________________________________________
> click mailing list
> click at amsterdam.lcs.mit.edu
> https://amsterdam.lcs.mit.edu/mailman/listinfo/click