[ih] Reporter query on the history of greater-than in quoted replies

[Greg Skinner]

Inclusion in a news reader is probably what made it _popular_. But there may
have been an _individual_ who came up with it, and it got taken up into the
news reader(s) after people saw it and started copying it.
So probably the place to check now is to go into Usenet archives (assuming
they are still online, I lost track of their comings and goings) and look for
instances, trying to track the behaviour backwards in time...

There are searchable Usenet archives available via Google Groups (groups.google.com). It helps to know the Usenet group in question. Searching for groups containing ‘net.mail' returned net.mail, net.mail.headers, and net.mail.msggroup. I recall this type of discussion took place in those groups.

—gregbo

[John Levine]

Inclusion in a news reader is probably what made it _popular_. But there may
have been an _individual_ who came up with it, and it got taken up into the
news reader(s) after people saw it and started copying it.

I'd guess it came via the >From convention in unix format mailboxes to quote
lines that would otherwise look like messages separators.

Perhaps one of the early unix guys will remember why they did it that way.

[Dave Crocker]

Post by John Levine

Inclusion in a news reader is probably what made it _popular_. But there may
have been an _individual_ who came up with it, and it got taken up into the
news reader(s) after people saw it and started copying it.

I'd guess it came via the >From convention in unix format mailboxes to quote
lines that would otherwise look like messages separators.

Allman introduced the ">" quoting of From at the beginning of a line in
in delivermail, I believe, but certainly by the time of sendmail. As
you suggest, it was to prevent the string from being interpreted as a
start of message.

Post by John Levine
Perhaps one of the early unix guys will remember why they did it that way.

Might have been unix, but more likely was the Tenex msg, which was the
first program to have a 'forward' command.

d/

[John R. Levine]

Post by John Levine
I'd guess it came via the >From convention in unix format mailboxes to
quote lines that would otherwise look like messages separators.

Allman introduced the ">" quoting of From at the beginning of a line in in
delivermail, I believe, but certainly by the time of sendmail. As you
suggest, it was to prevent the string from being interpreted as a start of
message.

I think it has to be earlier than that. We had mailboxes with "From "
separators on timeshared unix systems in the 1970s. If I have a chance
I'll look at some of the old source code.

Regards,
John Levine, ***@iecc.com, Primary Perpetrator of "The Internet for Dummies",
Please consider the environment before reading this e-mail. https://jl.ly

[Dave Crocker]

Post by John R. Levine

Post by John Levine
I'd guess it came via the >From convention in unix format mailboxes to
quote lines that would otherwise look like messages separators.

Allman introduced the ">" quoting of From at the beginning of a line in in
delivermail, I believe, but certainly by the time of sendmail. As you
suggest, it was to prevent the string from being interpreted as a start of
message.

I think it has to be earlier than that. We had mailboxes with "From "
separators on timeshared unix systems in the 1970s. If I have a chance
I'll look at some of the old source code.

As I wrote, I was referring to the angle-bracket quoting of 'from' at
the start of a line, to keep it from being interpreted as
start-of-message.

I don't recall quoting being done by the native Unix mail command, which
was all there was in Unix from the Labs.

(I think the actual use of mail as a separator was "<nl>from"...)

d/

[Guy Almes]

Dave et al.,
In the world of Internet History, this is a quite small matter, but

[1] it is an unending source of amusement and very occasional
frustration when one sees the '>' preceding the word "From" in web sites
and emails and even on text displayed on TV shows where, somewhere along
the line, some piece of software was making sure it was not
misinterpreted as the beginning of an email message. And very few
non-internet-nerds know what it's about.

[2] it also is a simple (and relatively harmless) example of how little
expedient decisions made *many* years ago, when combined with the need
for various forms of "compatibility", have consequences.

I suspect that long after the cockroaches have gone extinct, that '>'
will still be popping up.
-- Guy

Post by Dave Crocker

Post by John R. Levine

Post by John Levine
I'd guess it came via the >From convention in unix format mailboxes to
quote lines that would otherwise look like messages separators.

Allman introduced the ">" quoting of From at the beginning of a line in in
delivermail, I believe, but certainly by the time of sendmail. As you
suggest, it was to prevent the string from being interpreted as a start of
message.

I think it has to be earlier than that. We had mailboxes with "From "
separators on timeshared unix systems in the 1970s. If I have a chance
I'll look at some of the old source code.

As I wrote, I was referring to the angle-bracket quoting of 'from' at
the start of a line, to keep it from being interpreted as
start-of-message.
I don't recall quoting being done by the native Unix mail command, which
was all there was in Unix from the Labs.
(I think the actual use of mail as a separator was "<nl>from"...)
d/

[Dave Crocker]

Post by Guy Almes
[2] it also is a simple (and relatively harmless) example of how little
expedient decisions made *many* years ago, when combined with the need
for various forms of "compatibility", have consequences.

I probably stopped being surprised at its persistent 20 years ago. But
yes, it's really quite sad that it persists.

And yes, it is further proof that the evil we do lives long after us.
Interim at scale is the same as permanent.



d/

[Richard Bennett]

Internet has quite of few interim thingies. Interim is the fast path to standards.

Post by Dave Crocker

Post by Guy Almes
[2] it also is a simple (and relatively harmless) example of how little
expedient decisions made *many* years ago, when combined with the need
for various forms of "compatibility", have consequences.

I probably stopped being surprised at its persistent 20 years ago. But
yes, it's really quite sad that it persists.
And yes, it is further proof that the evil we do lives long after us.
Interim at scale is the same as permanent.
d/
--
Dave Crocker
Brandenburg InternetWorking
bbiw.net
_______
internet-history mailing list
http://mailman.postel.org/mailman/listinfo/internet-history

—
Richard Bennett
High Tech Forum <http://hightechforum.org/> Founder
Ethernet & Wi-Fi standards co-creator

Internet Policy Consultant

[Brian E Carpenter]

Post by Dave Crocker

Post by Guy Almes
[2] it also is a simple (and relatively harmless) example of how little
expedient decisions made *many* years ago, when combined with the need
for various forms of "compatibility", have consequences.

I probably stopped being surprised at its persistent 20 years ago. But
yes, it's really quite sad that it persists.
From my experience it's not only persistent but also necessary, resulting

from the continued use of Unix mailbox
fromat in all kinds of mail UA implementations.

Post by Dave Crocker
From Brian

[Dave Crocker]

Post by Brian E Carpenter

Post by Dave Crocker
From my experience it's not only persistent but also necessary, resulting

from the continued use of Unix mailbox
fromat in all kinds of mail UA implementations.

It represents a basic failure to keep a local hack local. It's fine if
the local implementation needs to do something special to preserve data
integrity. It is NOT fine for that modification of the data to escape
into the wild.

While the particular example is benign and even a bit amusing, it
highlights the need for implementation efforts that make changes made to
incoming data to reverse them on outgoing data.

d/

d/

[Brian E Carpenter]

Post by Dave Crocker

Post by Brian E Carpenter

Post by Dave Crocker
From my experience it's not only persistent but also necessary, resulting

from the continued use of Unix mailbox
fromat in all kinds of mail UA implementations.

(That was a small experiment to check that at least my UA
is very specifically looking for the string "From ".)

Post by Dave Crocker
It represents a basic failure to keep a local hack local. It's fine if
the local implementation needs to do something special to preserve data
integrity. It is NOT fine for that modification of the data to escape
into the wild.
While the particular example is benign and even a bit amusing, it
highlights the need for implementation efforts that make changes made to
incoming data to reverse them on outgoing data.

The underlying computer science error is perhaps a failure to distinguish
between data and metadata. But on the other hand, consider how important
it still was in the 1970s to economise on memory and instructions: we
all used little tricks of this kind in those days.

Brian

[John R. Levine]

I don't recall quoting being done by the native Unix mail command, which was
all there was in Unix from the Labs.

if (isfrom(line))
fputs(">", tmpf);

The v6 mail program is quite different and doesn't do that. The v7 mail
pogram knew about uucp and shipped off any message with a ! in the
recipient address.

Regards,
John Levine, ***@iecc.com, Primary Perpetrator of "The Internet for Dummies",
Please consider the environment before reading this e-mail. https://jl.ly

[Dave Crocker]

And, of course, 1979 is extremely late, in terms of the current topic.

d/

[John R. Levine]

Post by Dave Crocker
And, of course, 1979 is extremely late, in terms of the current topic.

I suppose. Berkeley didn't get Unix running on a Vax before 1979 and
PDP-11 2BSD was also in 1979 so it's hard to see how Allman's stuff could
have been earlier than that.

R's,
John

[Clem cole]

Mumble. I fear V7 is a bit too late. Delivermail (which predates Eric sendmail) was written by Kurt Shoens at UCB for v6 and I pretty sure Kurt did it. But I don’t think the idea was Kurt’s. IIRC it’s on the original BSD (aka 1BSD) tape. Which was released in 78.


That said the >From stuff was part of AT&T mbox format which I’m fairly sure is much earlier. There were a number of programs called mail that bounced around with Fifth & Sixth Edition in my memory. Where they all came is very fuzzy. Some of them like Rand changed the mailbox format. But others did not, particularly the ATT ones because of support for things like UUCP, Datakit et al.

My >>memory<< was use of mbox was driven by ATT not outside like ArpaNet needs (which drove Rand).


FWIW: CMU Unix in those days was hash of things from Bell and a lot of local hacks. The Unix boxes could not send email outside each other and since we had arpanet email on the pdp10s, no one tended build things like mail forwarding.

So the Unix systems I remember were not hacked beyond Unix style needs and I think we used the mailers as we got them but I don’t trust my memory on this. My thinking is we used something in EE at least that Ted brought with him from ATT Summit that was probably apart of UNIX/TS around 1976.


Sent from my PDP-7 Running UNIX V0 expect things to be almost but not quite.

Post by John R. Levine

I don't recall quoting being done by the native Unix mail command, which was
all there was in Unix from the Labs.

if (isfrom(line))
fputs(">", tmpf);
The v6 mail program is quite different and doesn't do that. The v7 mail
pogram knew about uucp and shipped off any message with a ! in the
recipient address.
Regards,
Please consider the environment before reading this e-mail. https://jl.ly
_______
internet-history mailing list
http://mailman.postel.org/mailman/listinfo/internet-history

[John R. Levine]

Post by Clem cole
My >>memory<< was use of mbox was driven by ATT not outside like ArpaNet needs (which drove Rand).

Oh, definitely. The earliest AT&T Unix I used in about 1975 had mbox
mailboxes which are clearly a hack someone invented in about two minutes
when he realized that he'd save a lot of disk space by putting all the
mail in one file rather than a file per message.

I don't remember whether they did >From quoting before the v7 mail program
I found.

R's,
John

[Dave Crocker]

Post by Clem cole
Mumble. I fear V7 is a bit too late. Delivermail (which predates Eric sendmail) was written by Kurt Shoens at UCB for v6 and I pretty sure Kurt did it. But I don’t think the idea was Kurt’s. IIRC it’s on the original BSD (aka 1BSD) tape. Which was released in 78.

From a paper that Allman co-authored:

"Sendmail started out as delivermail, written by
Eric Allman, then a graduate student and staff member
at the University of California at Berkeley. Delivermail
solved the problem of routing mail between three
different networks running on the Berkeley campus at
the time: the ARPAnet, UUCP, and BerkNet. The first
public version was distributed in 1979 as part of the
Fourth Berkeley Software Distribution (4BSD) and
later as part of 4.1BSD [Allm85]."

Sendmail Evolution: 8.10 and Beyond
http://www.sendmail.org/~gshapiro/Sendmail-8.10.Paper.pdf


That is, Eric wrote both delivermail and sendmail. As I recall,
delivermail was essentially a switching mechanism, cobbling together
existing email functions. sendmail was more integrated.

d/

[Steffen Nurpmeso]

Clem cole wrote in <08AB32A8-8AC3-42E8-8BBD-***@ccc.com>:
|Mumble. I fear V7 is a bit too late. Delivermail (which predates \
|Eric sendmail) was written by Kurt Shoens at UCB for v6 and I pretty \
|sure Kurt did it. But I don’t think the idea was Kurt’s. IIRC it’s \
|on the original BSD (aka 1BSD) tape. Which was released in 78.
|
|That said the >From stuff was part of AT&T mbox format which I’m fairly \
|sure is much earlier. There were a number of programs called mail \

V5 usr/src/s2/mail.c uses append("\nFrom ",buf) and creates
entries which are still valid according to POSIX (mailx), i think.

|that bounced around with Fifth & Sixth Edition in my memory. Where \
|they all came is very fuzzy. Some of them like Rand changed the \
|mailbox format. But others did not, particularly the ATT ones because \
|of support for things like UUCP, Datakit et al.
|My >>memory<< was use of mbox was driven by ATT not outside like ArpaNet \
|needs (which drove Rand).
|
|FWIW: CMU Unix in those days was hash of things from Bell and a lot \
|of local hacks. The Unix boxes could not send email outside each other \
|and since we had arpanet email on the pdp10s, no one tended build things \
|like mail forwarding.
|
|So the Unix systems I remember were not hacked beyond Unix style needs \
|and I think we used the mailers as we got them but I don’t trust my \
|memory on this. My thinking is we used something in EE at least that \
|Ted brought with him from ATT Summit that was probably apart of UNIX/TS \
|around 1976.

This is what i thought: "Unix style needs".
V5 did not have any `reply' command and "Save?"d "mailbox" message
bytes were simply appended to the "mbox" file. (Reading the
"mailbox" simply wrote the content to file descriptor 1.)

I do not know code earlier than V7 which performed ">From "
quoting to avoid misinterpreting these boundaries, too.

|Sent from my PDP-7 Running UNIX V0 expect things to be almost but not \
|quite.
|
|> On Jul 22, 2018, at 4:22 PM, John R. Levine <***@iecc.com> wrote:
|>
|>> On Sun, 22 Jul 2018, Dave Crocker wrote:
|>> I don't recall quoting being done by the native Unix mail command, \
|>> which was
|>> all there was in Unix from the Labs.
|>
|>> From the source code to mail.c in v7 Unix, dated May 1979:
|>
|> if (isfrom(line))
|> fputs(">", tmpf);
|>
|> The v6 mail program is quite different and doesn't do that. The v7 mail
|> pogram knew about uucp and shipped off any message with a ! in the
|> recipient address.
|>
|> Regards,
|> John Levine, ***@iecc.com, Primary Perpetrator of "The Internet \
|> for Dummies",
|> Please consider the environment before reading this e-mail. https://jl.ly

--steffen
|
|Der Kragenbaer, The moon bear,
|der holt sich munter he cheerfully and one by one
|einen nach dem anderen runter wa.ks himself off
|(By Robert Gernhardt)

[Dave Crocker]

Post by Steffen Nurpmeso
V5 did not have any `reply' command and "Save?"d "mailbox" message

The first reply command (called answer in that system) was the BBN Tenex
(on DEC-PDP-10) MSG system written by John Vittal. I believe it was
also the first with forward.

There was no need for a quoting character until these functions started
getting used. Mid '70s.

d/

[Dave Crocker]

Yes both Answer (now called reply) and Forward were invented in MSG,
sometime between 1974 (when Vittal got Yonke's BANANARD software) and
1976 (by which time MSG was a fully fledged MHS).

<quibble>

MSG was only an MUA and never did transport. So it was specifically not
an MHS.

It invoked SNDMSG to send mail. And it relied on the existing
receiver-side software to get mail into the inbox.

</quibble>


d/

[Craig Partridge]

Completely agree -- brain failure -- typed MHS when I meant UA, said to
myself "go fix" and failed to do so.
Sorry to have forced a correction!

Post by Dave Crocker

Yes both Answer (now called reply) and Forward were invented in MSG,
sometime between 1974 (when Vittal got Yonke's BANANARD software) and 1976
(by which time MSG was a fully fledged MHS).

<quibble>
MSG was only an MUA and never did transport. So it was specifically not
an MHS.
It invoked SNDMSG to send mail. And it relied on the existing
receiver-side software to get mail into the inbox.
</quibble>
d/
--
Dave Crocker
Brandenburg InternetWorking
bbiw.net

[Jack Haverty]

My recollection is that the ">" convention began very early in the
history, probably 1972 +- a year or so.

At that time, I used a PDP-10 running ITS (MIT-DM). Incoming mail was
appended to a file in your directory. Outgoing mail was usually
prepared in an editor (Emacs or TECO) and dropped into a directory where
the mailer could pick it up and send it.

Since there were no ARPANET standards for forwarding, replying,
redirecting, or otherwise dealing with a message that had already been
sent and received, people invented their own conventions.

E.G., in replying or forwarding, several conventions were popular.
Sometimes the "old" message was indented. Sometimes it had each line
prepended with some character such as - + = | >. Sometimes it was just
appended to the "new" message, with some kind of separator (line of ====
for example). Sometimes your new text was interspersed inline with the
old. You create your new message, possibly containing (pieces of) one
or more older messages from your inbox, and wrote the new message as a
file accessible to the mailer for delivery.

Various TECO and Emacs macros were quickly written to make this easier.
E.G., a macro could easily take the old message in your edit buffer and
put a > at the beginning of each line. I imagine the ">" approach won
out when someone eventually chose it and locked its use into code inside
a "mail program" that was less easily modifiable by mere users.

I recall struggling with this situation when I was building a
"communications daemon" that attempted to deal sensibly with forwarded
and replied-to incoming messages. There were simply too many personal
conventions being used to make it possible to write code to figure out
exactly what was going on in all the structure of an incoming message
that might contain parts of prior messages.

The headers were especially gnarly - e.g., trying to figure out who
originally sent a message that had been forwarded.

That led to the creation of the "MESSAGE-ID" header field, in an attempt
to get a "hook" into the protocol to make such things possible in the
future. The "new protocol" was supposed to take care of all this. We
never got there....

/Jack Haverty

[Lars Brinkhoff]

Post by Jack Haverty
My recollection is that the ">" convention began very early in the
history, probably 1972 +- a year or so. At that time, I used a PDP-10
running ITS (MIT-DM). Incoming mail was appended to a file in your
directory. Outgoing mail was usually prepared in an editor (Emacs or
TECO)

EMACS first appeared in 1976. I actually have EMACS version 24 running;
it's from November that year. It can call out to the RMAIL TECO package
for hanling email. RMAIL predates EMACS by a few years or so.

[John Levine]

Post by Dave Crocker
That is, Eric wrote both delivermail and sendmail. As I recall,
delivermail was essentially a switching mechanism, cobbling together
existing email functions. sendmail was more integrated.

I suppose we could ask Eric. In that era code and people were flowing
back and forth between Berkeley and Bell Labs. Since >From apparently
appeared in both places by 1979 it's anyone's guess which way it
initially went.

R's,
John