[Oberon] filesystem with directories ?

Thomas Frey thomas.frey at alumni.ethz.ch
Wed May 3 21:06:16 CEST 2006

> Hmmm...that's intersting.  So I take it AosHTTPServer
> assumes the "last" "." is the one not to convert?
No assumptions are made. The Server converts from the "/" path given
in the URL to the "." structure on AosFS, so there is no ambiguity.

> Anyway, that still doesn't cover what I was talking
> about with regards to user permissions.  Unless
> there's someway in AosHTTPServer to assign
> certain permissions to any file prefixed
> with my.folder.structure.
On Aos/Oberon there is only that much "permissions" as is programmed
into the FTP server (since it is the only program that uses
permissions), so this can be changed trivially.

> you can have such deeply nested file structures.
> There have been many times I've "drilled down"
> many levels deep only to find a single file,
> or worse an empty folder.
Then you maybe need to reorganize your folder structure (maybe to a
flat system if that works for you)
Many people seem to like searching for files or browsing for them. I
prefer to know where they are and have a mental log(n) access by an
organized folder structure. Tastes and work environments are

> Remember the example I gave of Leonardo.Text?
> Someone could structure that as:
> /Program Files/Leonardo/Documentation/Readme.Text
> but is that REALLY necessary?
Nothing is necessary. I never want to read a readme file anyways (even
though it can not always be avoided).

> Only if you make the user interface look like
> folders.  I often use "Prefix.*.Suffix" from
> the Documents.Panel.  It looks nothing like
> folders (to me anyway).  I find (pardon the
> pun) this a faster way to find stuff than
> "drilling down" in folders.  I like the fact
> that Plugin Oberon shows all the files in
> the "Directories" list as together in one
> list.

When I search for an assignment for some course at ETH for example, I
would have do a search for example for "ETH*"  to bringing me all
files that have to deal with ETH.

Which in my case results in about 20'000 different files. From there I
have to think of or know the search mask that results in the ones I am
interested in. If I know the search mask then I can of course skip any
Otherwise I "quickly" browse the 20'000 results to figure out if I
used "Courses" or "Vorlesungen" or "Lectures". From then I want to get
an overview about the courses (many instances for many years and
search for the newest ones) again browsing - this time only a few
thousand files - and "soon" I have found my desired list of files,
sorted by name they are easy to find because they appear grouped
because of the same prefix...(almost as if they were in a folder or
group or something the like) but then again I should not try a sort on
file dates or size or type, before the mask really only contains the
files I want (otherwise they are no longer grouped, because the groups
and filenames in a "."ed filename are not orthogonal but intermingled)

The same scenario with something like folders takes selecting 3-4
times out of folder lists of maybe 10 items each. I dont need to know
any names, i just need to recognize what i need. (all assuming a
reasonable folder structure)
Maybe this issue is reduced a bit if one only names files in one
language and has a much better memory than me. Myself I want my
filenames in the "native" language of the respective files, be it

> Again, from my own experience, many "file folders"
> could go away with a little forethought.
Sure, they all can go away but myself I dont want them to go away. I
would rather have "more" grouping features than less. Every strict
hierarchical structure can always be flattened into a prefix free code
(e.g. dotted filenames, but there is no limit in the number of
possible other encodings).
Personally I think it is better to keep all attributes (groupings,
permissions and even names) and contents orthogonal to each other
instead of compressing everything into the filename.
The trend on windows/Macintosh of course clearly goes away from
ordered structures with reasonable names towards almost anonymous
entities where file names are just (almost) useless formalities. Why
do we need grouping attributes (folders) or even filenames when we can
just search by content ?

> > If you want to get a feeling of how your own
> > directory structure could
> > look after a "dotification", you can run the
> > following commands:
> > Windows NT/XP: dir /s /b > preview.txt
> > Linux: find / > preview.txt
> > Then open preview.txt in a text editor and replace
> > "\" respectively "/" to "."
> > Then open preview.txt in a text editor and replace
> > "\" respectively "/" to "."
> And what would that really show?  There's a
> lot on my Windows box that's not part of MY
> directory structure.  (i.e. everything under
> C:\Program Files)  I typicially don't have
> folders more than 4 deep for things I set
> up myself.

Better separate the system and your data with two different
partitions, it gets confusing otherwise.
If I include the files on my system partition as well into the
statistics, it looks much worse (there are 38388 directories and
120719 files more on my windows system/program partition) I did not
include any system files because I am from a user's point of view not
interested in the content of my system/program partition, I don't
consider this as my data. I don't need to find things there normally
(unless I want to mess around in the system) and i also don't care
much about the ordering/naming there.


More information about the Oberon mailing list