Re: Longhorn Killer

[gerold]

On Tuesday 23 March 2004 15:22, Rogelio Serrano wrote:
> --- Raffael Herzog <herzog-lists@raffael.ch> wrote:
> On Mon, 22 Mar 2004 16:57:32 -0800 (PST), Rogelio Serrano
>
> <rogelio@smsglobal.net> wrote:
> > One of my basic problems is sorting and searching my emails. I have 10k
> > plus in my mail folders already. Searching seems so slow and its
> > difficult to search based on content. I think it goes for all my other
> > files too. I tried writing all emails to individual files and using sym
> > links but that was worse. It thrashed my hard disk like crazy.
> >
> > I realize now that its really inefficient to use a database as a
> > filesystem. Specially mysql or any other client server rdbms.
>
> Well, now you're contradicting yourself. DB based systems can solve
> exactly the problems you're referring to in the first paragraph I quoted.
> As an example (and to get even more off-topic ;): Look at Opera's M2. I'm
> managing the mails of three IMAP and two NNTP accounts with this thing,
> there must be tens of thousands of messages, not a single one of them
> stored locally on my computer -- and believe it or not: I find the message
> I'm looking for within seconds. Why? Because M2 is a DB based system.
> Instead of using folders, it uses views to organise mail. You can look at
> them as SQL-Views, i.e. prepared SELECT statements on your mail/news -- on
> *all* of it, no matter where it's stored. Combined with quick access to
> several filters and a quick-search facility (incremental search on what's
> currently displayed), this is a very efficient system.
>
> DBs are made for such tasks, the need for managing huge amounts of data
> and still quickly finding what you're looking for is actually the reason
> why they exist.
>
> cu,
>     Raffi
>
> [snip]
>
> I understand.
>
> So can everything on your filesystem be organized as views? Can the root
> filesystem be organized that way? I actually thought so when I started this
> thread.
>
> My original idea was to use a storage system on top of the filesystem.
> Since filesystems perform poorly with large numbers of really small files
> it would be logical to store small items together in bigger files. I
> thought that the notes data model as I understood it was simple enough for
> me to implement quickly.
>
>
>
> _______________________________________________
> Discuss-gnustep mailing list
> Discuss-gnustep@gnu.org
> http://mail.gnu.org/mailman/listinfo/discuss-gnustep

One filesystem that performs well with large numbers of small files is the 
Reiser  file system. It is currently in final testing, but the main advantage 
is that it is an atomic file system (either the operation succeeds or it 
fails entirely) which is the basics for any form of reliable database. A 
database could thus be built on top of this layer ?

More info at http://www.namesys.com/v4/v4.html

Have a good read,
-- 
Gerold Rupprecht
3, rue Louis-Curval
CH-1206 Geneva
Switzerland