mapping files to objects

[Mike McDonald]

>To: cwg@deepeddy.com
>Subject: Re: mapping files to objects 
>Date: Tue, 06 May 1997 16:50:03 -0700
>From: Kelly Murray <kem@franz.com>
>
>
>> > So who really believes a real or virtual working LispM can
>>  > and will be created?

  Well, since it feels like I've written about half of the 2.7MB of
mail, I think it should be pretty obvious that I do.

>> Depends on how you define a LispM.
>
>I guess the last 2.7mb of mail messages in which we've discussed exactly
>how we (and I) define a LispM hasn't been processed yet.

  Well, out of the 2.7MB of mail, there's probably something like 2.7M
different ideas of what a LispM is. :-) But I'm of the opinion that
there's not that big of a gulf left between the parties. (Maybe I just
have my rosey glasses on!) For example, the mail reader/POS debate.
Chris wants to work on a mail reader. Inorder for him to do that, he
needs a few things, namely an SMTP mail agent. Since no one is
currently signed up to write one, and even if someone was, it be a
while before it was ready, he's going to make the temporary compromise
of using the native Unix mail agent (aka sendmail). This compromise
means that he has to handle mbox formated files in the "short" term.
So he has to write some crufty file parsing code to get his objects in
memory. Ultimately, this objects will be based upon POS objects when
they become available. But since POS isn't available, he has to make
another short term compromise, he has to save/restore these mail
objects to a file. When POS is available, he can switch over to making
his message object based upon them and throw away his crufty file
code. Think of his file code as narrowly defined
checkpointing/restoration. Maybe when all of the other pieces are
available, Chris may choose not to throw away his file based code but
to keep it around in addition to the POS based stuff. Then when people
want to try out the PC LispM, they won't have to start by throwing out
their existing mail folders, including the 2.7MB of lispos mail. But
we don't have to make the decision today as to if or when we'll stop
supporting file based Email folders. That's something we can worry
about when we actually have a choice.

   So you see, I don't think we're that far apart. Chris is just
worring more about the interim steps than the final goal. I still
believe these two are reconcilable.

  Mike McDonald
  mikemac@engr.sgi.com