[9fans] another dusty upas question

[9fans] another dusty upas question

[erik quanstrom]

why does "upas/nedmail -c fubar3" create /mail/box/quanstro/fubar3/mbox
and not /mail/box/quanstro/fubar3?

	minooka; upas/nedmail -c fubar3
	creating new mbox: /mail/box/quanstro/fubar3/mbox

- erik

Re: [9fans] another dusty upas question

[Russ Cox]

> why does "upas/nedmail -c fubar3" create /mail/box/quanstro/fubar3/mbox
> and not /mail/box/quanstro/fubar3?

Because 'mail quanstro/fubar3' delivers to
/mail/box/quanstro/fubar3/mbox
and not /mail/box/quanstro/fubar3.

Russ

Re: [9fans] another dusty upas question

[erik quanstrom]

>> why does "upas/nedmail -c fubar3" create /mail/box/quanstro/fubar3/mbox
>> and not /mail/box/quanstro/fubar3?
>
> Because 'mail quanstro/fubar3' delivers to
> /mail/box/quanstro/fubar3/mbox
> and not /mail/box/quanstro/fubar3.

sure enough.  but why does it do that?

- erik

Re: [9fans] another dusty upas question

[Russ Cox]

>>> why does "upas/nedmail -c fubar3" create /mail/box/quanstro/fubar3/mbox
>>> and not /mail/box/quanstro/fubar3?
>>
>> Because 'mail quanstro/fubar3' delivers to
>> /mail/box/quanstro/fubar3/mbox
>> and not /mail/box/quanstro/fubar3.
>
> sure enough.  but why does it do that?

Because 'mail quanstro' delivers to /mail/box/quanstro/mbox
and not /mail/box/quanstro.

Russ

Re: [9fans] another dusty upas question

[erik quanstrom]

>>> Because 'mail quanstro/fubar3' delivers to
>>> /mail/box/quanstro/fubar3/mbox
>>> and not /mail/box/quanstro/fubar3.
>>
>> sure enough.  but why does it do that?
>
> Because 'mail quanstro' delivers to /mail/box/quanstro/mbox
> and not /mail/box/quanstro.

that makes sense.  thanks.

however, i'm not sure i fully understand the situation.  why does
	echo fu | upas/marshal -Fs bar quanstro/baz
deliver the copy to
	/mail/box/quanstro/(f/)?quanstro_baz
and not
	/mail/box/quanstro/baz/(f/)?quanstro
where it would land if i filed it after-the-fact with nedmail.

- erik

Re: [9fans] another dusty upas question

[Russ Cox]

All your previous questions were about mail delivery,
which I think is pretty consistent about not having
special cases.

> however, i'm not sure i fully understand the situation.  why does
> 	echo fu | upas/marshal -Fs bar quanstro/baz
> deliver the copy to
> 	/mail/box/quanstro/(f/)?quanstro_baz
> and not
> 	/mail/box/quanstro/baz/(f/)?quanstro
> where it would land if i filed it after-the-fact with nedmail.

Now you're into weird special cases.
How did you find the marshal -F flag, anyway?
It's not documented that I can see.
Maybe it should just be deleted.  What uses it?

Russ

Re: [9fans] another dusty upas question

[erik quanstrom]

>> however, i'm not sure i fully understand the situation.  why does
>> 	echo fu | upas/marshal -Fs bar quanstro/baz
>> deliver the copy to
>> 	/mail/box/quanstro/(f/)?quanstro_baz
>> and not
>> 	/mail/box/quanstro/baz/(f/)?quanstro
>> where it would land if i filed it after-the-fact with nedmail.
>
> Now you're into weird special cases.
> How did you find the marshal -F flag, anyway?
> It's not documented that I can see.
> Maybe it should just be deleted.  What uses it?

☺.  i found it by reading the source.  i was trying to track
down all the places where mailboxes are fidded in the upas
source.

it does seem very wierd and marshal does a shimmy to
pull it off.  it would seem the only way to file outgoing
& incoming mail to/from the same person in the same
file.

(offtopic: there are at least a few other crunchy bits.
several copies of cistrn?cmp by various names.  there's
also a reference to an rmail program in send/main.c
does it (did it) exist on plan 9?)

- erik

Re: [9fans] another dusty upas question

[Russ Cox]

> ☺.  i found it by reading the source.  i was trying to track
> down all the places where mailboxes are fidded in the upas
> source.
>
> it does seem very wierd and marshal does a shimmy to
> pull it off.  it would seem the only way to file outgoing
> & incoming mail to/from the same person in the same
> file.

ah, so the nedmail rf command invokes marshal -F.
i would be inclined to change -F to take the name
of the file, so that nedmail can be the sole arbiter of
what goes where.

> (offtopic: there are at least a few other crunchy bits.
> several copies of cistrn?cmp by various names.  there's

these could go.  they were in upas before they were in libc.

> also a reference to an rmail program in send/main.c
> does it (did it) exist on plan 9?)

i don't believe rmail ever existed on plan 9, but of course the
sources once compiled on unix too.  my understanding (which
may well be incomplete or incorrect) is that for much of the 1990s,
the bell labs mail gateways ran upas on irix machines,
and the binaries for those machines were generated by
compiling upas on plan 9, using vc and vl, but linking with
irix system call stubs.  perhaps behaving as "rmail" was still
necessary for that.  i also believe that's why vl has support for
writing elf binaries.  (but i might easily have the details wrong.)

russ

Re: [9fans] another dusty upas question

[erik quanstrom]

> ah, so the nedmail rf command invokes marshal -F.
> i would be inclined to change -F to take the name
> of the file, so that nedmail can be the sole arbiter of
> what goes where.

don't forget acme mail.

touching the poison upas is a thankless deal,
but at least one makes no progress.  and the
scars are cool.

- erik

Re: [9fans] another dusty upas question

[Russ Cox]

>> ah, so the nedmail rf command invokes marshal -F.
>> i would be inclined to change -F to take the name
>> of the file, so that nedmail can be the sole arbiter of
>> what goes where.
>
> don't forget acme mail.

the "pick a save file based on address" logic is unique
to nedmail.  acme mail has nothing like that.

russ

Re: [9fans] another dusty upas question

[erik quanstrom]

>> ☺.  i found it by reading the source.  i was trying to track
>> down all the places where mailboxes are fidded in the upas
>> source.
>>
>> it does seem very wierd and marshal does a shimmy to
>> pull it off.  it would seem the only way to file outgoing
>> & incoming mail to/from the same person in the same
>> file.
>
> ah, so the nedmail rf command invokes marshal -F.
> i would be inclined to change -F to take the name
> of the file, so that nedmail can be the sole arbiter of
> what goes where.

there's one problem with that.  marshal saves a copy of the
email after all the attachments have been made.

i can't replicate this easily with -C.

echo x | upas/marshal -C quanstro/quanstro -s y quanstro/z
Mail  from 'quanstro'
to 'quanstro/quanstro' alias `local!quanstro/quanstro'
failed with error 'unknown user'.

- erik

Re: [9fans] another dusty upas question

[Russ Cox]

> there's one problem with that.  marshal saves a copy of the
> email after all the attachments have been made.
>
> i can't replicate this easily with -C.
>
> echo x | upas/marshal -C quanstro/quanstro -s y quanstro/z
> Mail  from 'quanstro'
> to 'quanstro/quanstro' alias `local!quanstro/quanstro'
> failed with error 'unknown user'.

i don't understand.

right now nedmail invokes "marshal blah blah blah"
to send an r reply mail and "marshal -F blah blah blah"
to send an rf reply mail.  there is no reason it can't
invoke "marshal -F /mail/mbox/quanstro/f/rsc blah blah blah".
it just saved the message being replied to, so it knows the name of
the file.

russ

Re: [9fans] another dusty upas question

[erik quanstrom]

>> there's one problem with that.  marshal saves a copy of the
>> email after all the attachments have been made.
>>
>> i can't replicate this easily with -C.
>>
>> echo x | upas/marshal -C quanstro/quanstro -s y quanstro/z
>> Mail  from 'quanstro'
>> to 'quanstro/quanstro' alias `local!quanstro/quanstro'
>> failed with error 'unknown user'.
>
> i don't understand.
>
> right now nedmail invokes "marshal blah blah blah"
> to send an r reply mail and "marshal -F blah blah blah"
> to send an rf reply mail.  there is no reason it can't
> invoke "marshal -F /mail/mbox/quanstro/f/rsc blah blah blah".
> it just saved the message being replied to, so it knows the name of
> the file.

sorry.  i thought you were talking about removing the -F option
from marshal all together.

- erik