Re: IMAP server (was: Re: [Savannah-hackers] gnu-socketserver)

[Nic Ferrier]

>>> "Alain Magloire" <address@hidden> 27-Feb-01 2:23:54 AM >>>

>So doing something simple like:
>  msg.getLineNumber() ; // Fictuous example.
>Can be a killer, if the message was not dowloaded yet.
>My guess is JavaMail is trying to do a lot of stuff in the 
>background for you.  GNU Mailutils in a sense suffers 
>from this, but is less agressive or more dumb.

Ah no... they don't get off the hook that easy. There's a method in
Javamail (Message I think) that allows you to specify what will be
pre-fetched. AFAIK it's not used.

But if I don't call it I don't expect anything to come back... a
completly empty object. When I do something I expect to take the hit.
If I write a GUI client I would "do" everything at once (and probably
speicify what I wanted to pre-fetch) thus having one hit that does
everything.

But if I write another sort of client, an MTA delivery agent for
example, I don't want that hit in the same place. I want to control
it.

Having JavaMail do stuff like download the entire mail to de-b64 it
is utterly stupid.


>In essence they concluded that programming models trying 
>to make "remote objects" similar to programming local objects 
>are flawed ;-).

Absolutely. I don't think it's just Sun. Bad programmers want this
sort of stuff because they think it makes their lives easier (it
doesn't it just introduces complexities they don't understand). Good
programmers want this stuff because it's fun to implement. Companies
want this stuff because they get to sell a whole new wave of stuff to
people who only just finished paying for the last lot of stuff and
because it keeps the support subscriptions rolling in.


>Good idea, I will be curious to see the result, 
>because I presume that the working group faced 
>the same problems.  But with different goals and now 
>with field experience with JavaMail the conclusions
>may be different.

And slightly different field experience from the norm. Where I am
coming from is server side usage. I appreciate that may not be the
dominant area of usage for javamail but it does provide some
interesting insights.

In fact my main critisism of Javamail's designers and implementors is
that they are just way too client centric. They should have thought a
lot more about the other environments where javamail might run.


When we finish the oje mail implementation (and I'm in the last
throws of the roll out from hell with talk21 at the moment so I'm not
getting much time) I'll obviously let you see it... hopefully it will
be a lot faster than currently.

Hopefully you'll join us on any future JSR we manage to get going.

And when I have finished my website and got some info about the IMAP
server (and the code into CVS) I'll pop you a line.


Nic