Re: Improving Hurd

[Jan Atle Ramsli]

Jeff Bailey wrote:

> Perhaps you've noticed that there have been a fair number of new
> conributors.
I have noticed a significant improvement, and it may just have coincided
with the change to L4, but I don't think it's coincidence.
The idea of having an API where yo can mentally picture your message go
from here to there, and what state the system is in at each of those two
points, is attractive.
I think many people have been sitting on the fence like me, chewing on a
straw, and when it started to look like fun, they jumped off and joined
in.
> 
> Everyone who shows up seems to think they can solve the "big"
> problems.
When building a block device driver becomes a matter of sitting on the
interrupt vector and moving stuff between ports (in the sense of IN,OUT,
X, Y) and a buffer, kicking the device's rear end in xx_interrupt() or
similar in one end, and moving stuff from that buffer to and from queues
in xx_operations in the other, then, a block device is not a 'big'
problem anymore.

It is just a matter of expressing an ADT in terms of API.

The hairy stuff it supposed to be in one place.

And I see improvements, I must confess that it has been a while since I
have unpacked it, maybe I should take your response to mean that it is
time to have another look.

Atle