The discovery

[Jean-Francois Dive]

Hi Click team,

I just wanted to drop you an email to tell you that i am reading all the architecture of your work, read and read again (code etc...) and have only one word... brilliant. Really, i mean it. Totally great.

I've got a few questions about it:

--> Interraction with the linux TCP/IP stack: from what i can understand, most of the CPU time at the kernel level is used by the main thread processing the click tasks. It release it from time to time to allow to linux to still survive. (this is in standard, i did not read the SMP extentions). My main point is to know if it is still valid to think to use applications (network daemons) on top of the standard stack (i know i have to configure the elements to send some traffic to the stack). How will the performances be affected, in a general way ?

--> I understand that the dedicated channel of communication between the process level and the kernel modules is /proc . Is there any other possible hooks (to READ/WRITE informations to elements in the kernel ?) ?

--> Is there a plan or a good idea to have a possibility to integrate standard daemons with data's inside the click structures (using zebra directly for exemple)... i believe not, but maybe i missed another great idea ?

--> Is there a list of already developped elements or people starting to do so ?

Thanks for bringing this,

Cheers,

JeF
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://amsterdam.lcs.mit.edu/pipermail/click/attachments/20011024/d9b305ca/attachment.htm