Re: [Gnumed-devel] (no subject)

[Karsten Hilbert]

Please keep posting to the gnu.org list.

> I don't know why you are being so negative about it all. 
Oh, this is my second nature :-))

I do agree that it'd be nice to recover data that was typed in
before a client crash. I tried to explain a few times why I
think it is unsafe. I will try once more:

Before the crash I was connected to the databases on the server
at my office. After the crash - because I can't get the server
at the office going again - I connect to the server at my
house which happens to sync itself 3 times daily with my
office (remember, Postgres-R isn't released yet). Now, what is
my client expected to do with data I typed in before the crash
that it picks up from its local hard disk ? Commit it to the
server at my house ? You got to be kidding.

This isn't impossible to implement. It is just not easy and
will take quite some time (which people may take if they want
but they shouldn't hinder getting GnuMed into a useable state
otherwise by taking their time).

>   I don't think either you or me have tried to implement it, Karsten.
Because *I* chose not to spend time on it and for the reasons
mentioned above.

>   At work , we use a common Australian GP system which does:
>       1) restart with typed data if the computer happens to freeze for
> some reason and the reset button has to be pressed. When the system
> restarts, it detects an incompleted patient session and enquires wether
> to continue with the session. If the session is continued, it appears
> what was most recently typed a few seconds ago is recovered.
I bet you a beer that this system does not expect it's database
server to change under itself inbetween crashes and I bet that
it'll fail catastrophically should that happen. I may be
wrong. The better for it.

>       2) if you are editing with vi a program and your computer freezes or
> shuts down, when you reopen the file for re-editing , vi detects an
> existing swap file and requests whether you wish to continuing editing,.
> ignore it , or delete it. 
Where exactly does this relate to arbitrary distributed
database connection state ?

> You  seem to think that blanket decress are the end all of discussion,
Do I ?

> and I don't think this is a particular positive approach to sharing
> knowledge, which is what this is all about, isn't it?
It is also about getting GnuMed into a useable state in a
reasonable amount of time.

I am sorry for having strong opinions on some matters :-)

Karsten
-- 
GPG key ID E4071346 @ wwwkeys.pgp.net
E167 67FD A291 2BEA 73BD  4537 78B9 A9F9 E407 1346