RE: Three silly questions to the Hurd-Core
Some overly simple answers:
1) Where to start? First, as you have learned, there are not good documents
describing the Hurd. After reading the overviews you have little choice but
to jump in and start reading the program code. The best way I've found to
do that is to start with a question (any question will do) and try to track
down
an answer. You will likely get at best an incomplete answer. Be happy. You
have made progress. Then pick a different question and try to track down
an answer. Over time your questions will become more productive and your
answers
more complete. The phrase "code is the best documentation" is misleading.
As documentation, code is often poor, but at least it is accurate.
2) Translators are important, and, like anything that is truely
new, it is full of surprizes for the future. I think it is best described as
way
for a developer to create new features that look as if they had been build
into the operating system from the beginning. It makes possible elegant
solutions to problems that before were complex. Initially it will be used
to do some of the same old things, just differently. When we finally get
the idea, we'll be inspired to try things we had never considered before.
3) Hurd is built on top of Mach. Certainly true. Mach is a means to an end,
not a foundation. Some of the Hurd developers have become very frustrated
with some of the things Mach didn't do or didn't do well. Over time, if
Hurd begins to live up to its promise, those limitations will be addressed.
Perhaps Mach will be improved, perhaps it will be replaced. I don't think
it matters which. Once it works well enough, most of use will probably
just ignore it and focus on the new possibilities that Hurd offers.
-----Original Message-----
From: Zsombor Gergely [mailto:gergely@econ.core.hu]
Sent: Friday, March 10, 2000 4:55 PM
To: debian-hurd@lists.debian.org
Subject: Three silly questions to the Hurd-Core
Dear Hurd-Core guys,
I have 3 silly questions.
1) Reading "all" the docs available, I still can not find the rationale for
several things (for having a non-logged in user, the design of the
networking
layer(s), etc). I still can not build a bridge between the verbal (easy
readings)
and the "real" part (code and description for programmers: data types,
buckets and the hell knows whatever...).
It has been mentioned that code is the best documentation, but what
do you suggest to somebody, who hase no _real knowledge_ of programming,
but
has an understanding of the possible underlying logic? I have the Gnumach
and the hurd sources, but I can not figure out, where to start. I think I
should start somewhere earlier (structure of a generic C program), right?
Or is it feasible to go right into the stuff itself? Any ideas out there?
2) Discussing features of the Hurd with another guy, he asked me what the
real benefits of it were. I tried to explain the possibilities of the
translator
structure, but even I did not find myself too convincing. Since then, I
think
that the possible security features (capabilities - see earlier thread on
Eros
and another post on a french interview - [sorry for no exact references])
and
scalability of the Hurd is at least as important as the translator concept.
I am right? If it is so, readings should stress this, and these could be a
focal point of the Hurd "marketing" as well.
3) Hurd -- nowadays -- still relies heavily on the mach microkernel. What
chance does it (Hurd) have to get ported on another one? Is there another
candidate anyway, or this is just a theoretical possibility? If I understand
well, the kernel (even a microcernel) have crucial role in making good SMP
work. Can Mach offer a good foundation for the Hurd in this sense?
Reply to: