debram and debtags

[Erich Schubert]

Hi,

> My package is the `vim' to your `emacs', as it were.

That is a very high claim, i hope you are aware of that.
If you read our mails to debian-devel carefully, you might have noticed
that we said we are in need for a "tags consortium" and a tagging crew.

Our claim is not to have a good database of tags for the packages.
But to provide an interface to browse the tagged data.

In fact it should be quite easy for us to import your data into our
database, and provide an *alternate* view of your data.

After having a 3-min look at your orig.tar.gz, i think that your system
uses a fixed hierarchy for browsing (like aptitude does).
Our system is "deriving" the hierarchy from the tagged data itself,
trying to provide a good balance between a wide and a deep tree.

Using a fixed hierarchy "optimizes" the browsing hierarchy for the
designer of the structure - which can be very unintuitive to others!
Whereas our hierarchy will depend on the use of tags by every maintainer.

Its easy to find some examples for that: i would not expect "spam",
"mail filtering" and "mailbox service" below "mail transport", whereas
"mailbox handling" is below "Networking Mail".
"Mail User Agents" is not below "Network Clients and Servers" etc.

BTW: i have found some miscategorizations in your system. Ocaml is not
an "X11 programming" thing, but belongs to the ML family of functional
languages.
"xmms-jess" is under SDL, but it belongs to xmms. Another
missclassification; and one that our system does prevent.
It will classify "xmms-jess" as (somthing in the lines of)
"sound, visualization, SDL, xmms" So you can approach it from many
directions - "sound, visualization", "sound, SDL", "xmms, vis" etc.

And i don't think cross-references can make the navigation as easy.
While you will be able to find something that you do know, you might
get lost when searching for some unknown program.

Libraries have one big limitation we do not have: they can put data only
in one place. We can have entries at any place we like to.
Also their cataloges are a mostly linear structure (you want to avoid
having to access one card box, just to find the next card box to look
in... but that is just fine with computers)

Still i don't claim your project is a waste of effort. It is a valueable
addition to our project, since it provides data we do not yet have:
- A hierarchy of "balanced" tags for the packages
- Meta information about packages

We have three big contestants now for the "tag database":
- the current "grown" tags used by debtags right now
- your hierarchy
- the "trove" categorization used by sourceforge and freshmeat

All of them will need major modifications unfortunately. No idea which
needs the least, but i think we should start at "Trove".

Greetings,
Erich Schubert
--=20
   erich@(vitavonni.de|debian.org)    --    GPG Key ID: 4B3A135C    (o_
       The best things in life are free: Friendship and Love.       //\
    Zwei Freunde m=FCssen sich im Herzen =E4hneln, in allem anderen     V=
_/_
       k=F6nnen sie grundverschieden sein. --- Sully Prudhomme