a little (folish?) idea

[Martin P. Hellwig]

Hi Folks,

Now that this group is created I can post the less df'ish stuff...
I've had an idea, probably unrealistic and maybe just plain stupid but I 
still like your opinion on it because I'm not talented enough to learn 
good C (let alone good math) in a reasonable time so that I get to code 
it myself , please don't mistake me for the “visionary type of guy”. I 
can assure you that I'm not,  I even  feel kinda handicapped not to be 
able to code this stuff together myself to prove my hypotheses.
Well I hope this is enough disclaimer, now the actual subject  ;-)

It sprang to my mind that almost all configuration / logging files are 
“plain” text files ,in my opinion a big pro, text files are very 
flexible stuff and easy handling.

However on a daily bases I work (and learn about) SQL and  SQL db's.
And in the way got charmed about the way it works .
One sleepless night I wondered if it wouldn't be handy if I could create 
a file which is ordinary in the sense that you can edit it with vi or 
whatever tool of your choice, but behind the scene is actually a view on 
a database.
And I mean this quite literal in the sense that it's not like a cron job 
which checks if the files are still the same, but more in the way of a 
“server side” push.
So that after the write on the edited file , the changes (after an 
optional sanity check) are updated in the database.
Other views change and so do the other “view-textfiles” who depend on 
the same (shared) table/data.

I could imagine that this would easer up multi-host configuration / 
authentication / logging.
You'll probably need to create a “textview” (a piece of text which 
describes how the data would be placed (syntax) in the file) a dbview 
(a plain database view) and optional sanity check for the input/output 
to/from/via the textview / dbview.
Well that would be the administrative tasks, the real challenge would be 
to create a certain framework which makes this possible.
But this is far away from my of expertise to see what the implication 
would be on the OS to allow this. Perhaps these mechanism is already 
used today but I'm unaware of it (I do know the /etc/(s)pwd.db structure 
but this is not direct enough IMHO because it still depends on other 
specialised tools like vipw and pwd_mkdb). Or even more probably this is 
just a bunch of bs.
But if you find the time please explain me why.

Thanks for your time reading this.

--
mph