[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Styx Speed and Implementation

forsyth@plan9.cs.york.ac.uk wrote:

> if you'd like a cpu server (say) to access a device on your machine,
> you have the server import your machine's device into its name space.
> X11-style network graphics, for instance, in a Inferno or Plan 9 environment
> isn't a specialised operation: a cpu server imports a terminal's graphics
> devices into its name space; that's it.

this is what I thought -- but how is this sort of thing coordinated
under inferno? Does your CPU server announce the service of accepting
output devices for a certain set of processes it's willing to run?

for instance, how would you choose to implement a "telnet" type
application, say on a much faster machine with a large store of
programs. You dial the machine, and then send it a message asking for an
instance of sh with your terminal mounted in its /dev/cons... is there
some library support for such an 'inverted' client/server arrangement?
Also -- much more curious about how and where the border between tk
'commands' being sent and /dev/draw messages being sent occurs: can you
mount the tk portion of your kernel into a remote server, enabling
high-abstraction (i.e. low bandwidth) instructions back to your
terminal? Anyone know? it's been a few weeks since I've fiddled with the

-graydon <admin@multinet.net>

aw shucks I musta been tuned to the wrong frequency