About using the filesystem for everything

FROM: Kurt Skauen
DATE: 05/30/2001 13:38:10
SUBJECT: RE:  [Atheos-developer] A couple of questions for Kurt [OT]

 "James Darpinian" <> writes:

> I find it kind of funny that we are all here discussing how AtheOS should
> work and what _we_ want it to do when it is actually _Kurt's_ OS.  I've
> noticed Kurt doesn't usually take part in the discussions (and I don't blame
> him, it would be a full time job! :-) and I was wondering if he even pays
> attention to all the decisions we're "making" for him.  So, a couple of
> questions for Kurt:
> 
> Do you read every post to this list and consider all the points they make?
> Or do you skim the titles and read only the ones that look interesting?  Or
> do you pay no attention to the user-interface discussions at all?

I try to skim trough everything but I receive to much mails to be able
to fine-read everything much less respond to it. Beside we are closing
in on a deadline at work so I have had to prioritize that for a while
now and that will continue at least some weeks to come.

The first time AtheOS was mentioned at slashdot I tried to keep track
of everything said about AtheOS and I ended up growing so tired of it
all that it took at least 6 months until I managed to do anything
productive again. Now I try to filter things a bit more to avoid
getting totally OD'ed again. I do of course consider the point's made
tough.

> Have you already decided most of the issues we're discussing?  Or
> are you open to suggestions for most things?

I have plans for most aspects but it is on a very "fluffy" level.  I
have a vague overview of most things but I tend to focus very hard on
whatever I'm working on at the moment and never iron out the details
of other parts of the project. This is a necessity for me to be able
to keep things on a level that I can handle and not getting totally
overwelmed by the all the details involved in making a OS.

I can try to comment on some of the issues that have been brought up
here though. First off, having the OS rename files is absolutely not
an options. A file can be reference by a million things that is
totally outside the scope of the OS so there is no way the OS can
fixup all the references it whould break by renaming the file. Beside
I want to be able to give a file whatever name I might desire whether
"." is includes somewhere in that name or not. The file-type will be
stored as a mime-type in an attribute like in BeOS. Native programs
should always try to set a mime-type on files it creates while a
"sniffer" should go over the FS every now and then updating files
without a mime-type (I have been thinking about making a special
"index" in AFS that contains all files without a mime-type to make it
easy for the sniffer to find what files to update without resorting to
iterate over every file in the entire file system).

AtheOS will have a "regular" FS. Hierarchial file systems seems to
work quite well. Be tried to replace the FS with a database and that
failed. They ended up with an attributed/indexed FS which I think is a
nice middle-way. Things doesn't neccessarily have to be bad just
because they have been used for a while. Some times tings last because
they work well.

I was only intending to use the current UNIX directories for ported
UNIX applications (something you generally won't bother with at all
until you open a terminal window). For native AtheOS applications I
want a separate directory tree with one app in each directory. Again I
don't iron out the details of something that is not the current focus
point.  This will be a focus point when the desktop manager are at a
level that makes it possible to start "random" apps from it.


Like I have mentioned in earlier posts already there will be standard
APIs to retrieve info about running apps, open windows, mime->app
mappings, etc etc that will be used by the desktop manager and that
will be available to everyone that might need's it. An icon will be
asocciated with each window and used in the ALT-TAB menu and the "task
bar". The icon will of course also be retrievable through the standard
API so any application can get to them.


-- 
Kurt Skauen. ( http://web.archive.org/web/20031125103357/http://www.atheos.cx/ )