Date:    Thu, 04 Nov 1999 14:46:53 +1100
To:      Gernot Heiser <gernot@cse.unsw.edu.au>
cc:      cs9242-list@cse.unsw.EDU.AU
From:    Vince <vincej@cse.unsw.EDU.AU>
Subject: Re: 9242 informal survey

Hi all,

> Should provide more consultation for project design issues (can learn
> more about advantages and disadvantages). 

This is surprising.... (I dont know about John's experience but speaking
for myself)

The consultations went a lot like  this...

If there wasnt a milestone due that week there were only a few people in
the lab... Those that were there (and asked questions) tended to ask
'designy' questions.  Most of the design question tended to be about pager
and filesys stuff.  comparisons of PT structures (usually inverted vs
heirechical) was quite common. However in these 'off' weeks sometimes
there was litereally nobody around in the consult times.

And if there was a milestone due, it tended to be monopolized by 

why is this IPC screwing up?...

which was always masively painful to try and figure out what was going on
considering the huge amounts of places you can stuff up a system from, and
that it is practically impossible to understand one of these projects
by just glancing at a few files...(especially in later weeks)

Apart from random bug-squashing (which was quite often only partially
successful) only a few students actually made use of me when I was there.
Students never seemed sure of when I would be in the lab (dont know why it
was on the web page). Basically if people didn't ask they didn't get
help... And quite a few never asked, especially about design issues.

In my opionion the consults were very useful for all in the begining of
session when people were still figuring out L4 as most people made use of
the consults as 'I dont understand whats going on here can you explain it
to me'... After the disk driver milestone most people seemed to understand
what was going on (ex_reg excluded see below) a lot of the problems after
that tended to be quite bizare. 

The consultations could probably be cut back in later weeks.. 

One other note...

The manual entry for lthread_ex_reg should have the following added in
huge letters of fire...

LTHREAD_EX_REG() CAN ONLY BE USED ON A THREAD IN THE SAME ADDRESS SPACE AS
THE CALLER. 

this is made worse by the fact theres no return value from ex_reg so no
error checking cant quickly be done by the caller (so students never
realise it isnt working at all)

Vince