Hybrid Computer-tab machine data centers?

[h***@bbs.cpcn.com]

In a library, I found a late 1950s IBM proposal to
install a 705-III computer system. Included with
the proposal was proposed system operation which included
considerable support work with traditional tabulating
machines. In other words, some files would be maintained
on punch cards alongside the computer and processed through
tab gear such as the collator, sorter, and tabulator. Only
the really complicated work would be fed into the computer
and processed to/from tape files.

The computer had no disks. However, it would've provided
an "on-line" inquiry function by suspending an ongoing
program, loading in a special search program, doing
a semi-indexed sequential search on tape files (that were
always mounted), and typing the results on the console printer.
Response time was estimated to be in minutes.

I believe they recommended ten tape drives.

Anyway, it appears back in the 1950s and early 1960s
that tab machines play a vital role in data center
operations working with electronic computers.

I wonder if that was a marketing decision by IBM so as to
keep older tab machines still in service earning rent, or
was necessary due to the limited I/O capability of electronic
computers. Or maybe the electronic computer could only do
so much work and the rest had to still be done the old way.

Remember, until the 1401 came out, card readers and printers
for IBM mainframes were terribly slow--reading and printing
150 lines a minute. IBM simply modified its existing tab
machines to serve as I/O units. The computer programs
didn't format data, rather it was done by plugboards.
The early machines allowed direct tape-to-tab and tab-to-tape
copying to process card files offline while the mainframe was
doing something else.

IIRC the proposal, a C/E would visit daily to power up the
machine and ensure all tubes were working and replace any
burned out.

The proposal itself was typewritten, using a carbon film
ribbon and proportional type. By 1950s standards it
looked good, but it was still obviously typewritten and
there was an occasional character that hit the line wrong.

[Micheal H. McCabe]

Post by h***@bbs.cpcn.com
In a library, I found a late 1950s IBM proposal to
install a 705-III computer system. Included with
the proposal was proposed system operation which included
considerable support work with traditional tabulating
machines. In other words, some files would be maintained
on punch cards alongside the computer and processed through
tab gear such as the collator, sorter, and tabulator. Only
the really complicated work would be fed into the computer
and processed to/from tape files.
The computer had no disks. However, it would've provided
an "on-line" inquiry function by suspending an ongoing
program, loading in a special search program, doing
a semi-indexed sequential search on tape files (that were
always mounted), and typing the results on the console printer.
Response time was estimated to be in minutes.
I believe they recommended ten tape drives.
Anyway, it appears back in the 1950s and early 1960s
that tab machines play a vital role in data center
operations working with electronic computers.
I wonder if that was a marketing decision by IBM so as to
keep older tab machines still in service earning rent, or
was necessary due to the limited I/O capability of electronic
computers. Or maybe the electronic computer could only do
so much work and the rest had to still be done the old way.
Remember, until the 1401 came out, card readers and printers
for IBM mainframes were terribly slow--reading and printing
150 lines a minute. IBM simply modified its existing tab
machines to serve as I/O units. The computer programs
didn't format data, rather it was done by plugboards.
The early machines allowed direct tape-to-tab and tab-to-tape
copying to process card files offline while the mainframe was
doing something else.
IIRC the proposal, a C/E would visit daily to power up the
machine and ensure all tubes were working and replace any
burned out.
The proposal itself was typewritten, using a carbon film
ribbon and proportional type. By 1950s standards it
looked good, but it was still obviously typewritten and
there was an occasional character that hit the line wrong.

In the late 1970's and early 1980's I used a UNIVAC 90/60 in both batch and
time-sharing modes. The student keypunch room where I punched most of my
FORTRAN and COBOL programs had an 084 sorter, a collator whose number I
can't recall, a card-to-printer accounting machine for listing card decks,
an interpreter, and other ancillary equipment for handling cards and paper
(burster, decollator, binding equipment, etc.)

At the time, students were encouraged to do as much with the tab equipment
as possible to minimize the workload on the computer.

One limiting factor was simply the number of terminals that worked at any
given moment. When I started, the standard terminal was a teletype model 33
ASR. Several teletypes were limited to "off-line" use for the production of
paper tapes. You would punch a tape with your BASIC (or FASTFOR) program
and wait for an "online" terminal to become available. Once an "online"
teletype was available, you grabbed it, did a conversational signon, entered
the language processor, then transmitted your tape. To my way of thinking,
this was a mix of both conversational and batch computing. I often punched
the results on tape as well, that way I could reproduce the output (on an
offline TTY) without having to run the program again.

Later (because of maintenance issues?), the model 33 teletypes were replaced
with model 43 teletypes. The model 43 TTYs didn't have the punch or reader,
so the offline TTYs were eliminated. At about the same time, we started
getting a few "glass teletypes" like the ADM-3 and ADM-5 CRT terminals. We
had these for less than a year before we started getting microcomputers. The
microcomputers replaced the mainframe for BASIC programming and early
systems courses.

The UNIVAC was in use until they replaced it with a VAX 11/780 in 1984.

Oh, one other gizmo we had in the room with all the tab equipment was a
card-to-tape system made by Mohawk Data Systems. This allowed you to submit
your batch programs and data on mag-tape instead of punch cards. The system
operator's never liked to mount student tapes, so I rarely used the
card-to-tape system. I never figured that out, since they never bitched
about card jobs.

Some other machines we had access to included a hacked 129 keypunch that
would put sequence numbers in your card decks automagically, and a real
abortion that consisted of a Hazaltine CRT terminal hooked up to some sort
of single board computer and an old model 24 keypunch. You could key and
edit text on the CRT, when completed you hit XMIT on the terminal and the
SBC punched a deck of cards on the keypunch. The model 24 didn't print on
the cards, so the next stop was the interpreter, then the accounting
machine. Finally, after a desk check, you could submit your cards to the
operator for batch processing. Typical batch turn-around during the day was
a little over an hour -- not really all that bad from what I've heard.

This was all at Edinboro University of PA from around 1978 to 1984.

[h***@bbs.cpcn.com]

In the late 1970's and early 1980's ...
At the time, students were encouraged to do as much with the tab equipment
as possible to minimize the workload on the computer.
From your report and others, my guess is that the cost of maintaining

old electro-mechanical gear was still less than electronics in those
days. We take cheap computing for granted these days and mechanical
devices are very expensive, especially the heavy duty IBM tab units.

One limiting factor was simply the number of terminals that worked at any
given moment. When I started, the standard terminal was a teletype model 33
ASR. Several teletypes were limited to "off-line" use for the production of
paper tapes. You would punch a tape with your BASIC (or FASTFOR) program
and wait for an "online" terminal to become available. ...

That too was common. It wasn't so much the cost of a terminal,
but rather the cost of the line and port, and CPU cycles of the
computer. The computer could only handle so much online traffic
and connected users so users were forced to be efficient about
it--as you were by pre-punching your program. (We did that as well.)

Later (because of maintenance issues?), the model 33 teletypes were replaced
with model 43 teletypes. The model 43 TTYs didn't have the punch or reader,
so the offline TTYs were eliminated.

Did you get more model 43s to replace the offline units or did
they run faster?

Oh, one other gizmo we had in the room with all the tab equipment was a
card-to-tape system made by Mohawk Data Systems. This allowed you to submit
your batch programs and data on mag-tape instead of punch cards. The system
operator's never liked to mount student tapes, so I rarely used the
card-to-tape system. I never figured that out, since they never bitched
about card jobs.

If the mainframe did not have a spooler, then tape would've been
much faster--IF--the tape was a reasonable volume. But I suspect
the system did have a spooler and most student jobs would be pretty
small. In that case, it seems foolish to use a data converter like
the Mohawk.

They were generally used in data entry operations as key-to-tape
rather than key punch.

[Micheal H. McCabe]

Post by h***@bbs.cpcn.com

In the late 1970's and early 1980's ...
At the time, students were encouraged to do as much with the tab equipment
as possible to minimize the workload on the computer.
From your report and others, my guess is that the cost of maintaining

old electro-mechanical gear was still less than electronics in those
days. We take cheap computing for granted these days and mechanical
devices are very expensive, especially the heavy duty IBM tab units.

One limiting factor was simply the number of terminals that worked at any
given moment. When I started, the standard terminal was a teletype model 33
ASR. Several teletypes were limited to "off-line" use for the production of
paper tapes. You would punch a tape with your BASIC (or FASTFOR) program
and wait for an "online" terminal to become available. ...

That too was common. It wasn't so much the cost of a terminal,
but rather the cost of the line and port, and CPU cycles of the
computer. The computer could only handle so much online traffic
and connected users so users were forced to be efficient about
it--as you were by pre-punching your program. (We did that as well.)

Later (because of maintenance issues?), the model 33 teletypes were replaced
with model 43 teletypes. The model 43 TTYs didn't have the punch or reader,
so the offline TTYs were eliminated.

Did you get more model 43s to replace the offline units or did
they run faster?

Yes, we ended up with more model 43s than we had with the online/offline
model 33's. The 43s ran at 300 baud, whereas the 33s ran at 110 baud.

Post by h***@bbs.cpcn.com

Oh, one other gizmo we had in the room with all the tab equipment was a
card-to-tape system made by Mohawk Data Systems. This allowed you to submit
your batch programs and data on mag-tape instead of punch cards. The system
operator's never liked to mount student tapes, so I rarely used the
card-to-tape system. I never figured that out, since they never bitched
about card jobs.

If the mainframe did not have a spooler, then tape would've been
much faster--IF--the tape was a reasonable volume. But I suspect
the system did have a spooler and most student jobs would be pretty
small. In that case, it seems foolish to use a data converter like
the Mohawk.
They were generally used in data entry operations as key-to-tape
rather than key punch.

As I recall, under VS/9 on the Univac, all I/O was spooled. That might
explain
the reluctance for the operator to mount a tape for the student jobs.

[rpl]

Post by h***@bbs.cpcn.com

In the late 1970's and early 1980's ...
At the time, students were encouraged to do as much with the tab equipment
as possible to minimize the workload on the computer.
From your report and others, my guess is that the cost of maintaining

old electro-mechanical gear was still less than electronics in those
days. We take cheap computing for granted these days and mechanical
devices are very expensive, especially the heavy duty IBM tab units.

One limiting factor was simply the number of terminals that worked at any
given moment. When I started, the standard terminal was a teletype model 33
ASR. Several teletypes were limited to "off-line" use for the production of
paper tapes. You would punch a tape with your BASIC (or FASTFOR) program
and wait for an "online" terminal to become available. ...

That too was common. It wasn't so much the cost of a terminal,
but rather the cost of the line and port, and CPU cycles of the
computer. The computer could only handle so much online traffic
and connected users so users were forced to be efficient about
it--as you were by pre-punching your program. (We did that as well.)

Later (because of maintenance issues?), the model 33 teletypes were replaced
with model 43 teletypes. The model 43 TTYs didn't have the punch or reader,
so the offline TTYs were eliminated.

Did you get more model 43s to replace the offline units or did
they run faster?

Oh, one other gizmo we had in the room with all the tab equipment was a
card-to-tape system made by Mohawk Data Systems. This allowed you to submit
your batch programs and data on mag-tape instead of punch cards. The system
operator's never liked to mount student tapes, so I rarely used the
card-to-tape system. I never figured that out, since they never bitched
about card jobs.

For small jobs, tapes are bulkier and require more actual
care-and-feeding than a little deck. Imagine the work required to
mount/dismount/keep-track-of 100 tapes as opposed to 100 1" thick card
decks which could just be stacked up sequentially in the hopper. By the
time you've taken the ring off the tape and opened the tape drive the
card job's already been inputted, rubber band wound and tossed back onto
the "pickup" table.

Post by h***@bbs.cpcn.com
If the mainframe did not have a spooler, then tape would've been
much faster--IF--the tape was a reasonable volume. But I suspect
the system did have a spooler and most student jobs would be pretty
small.
In that case, it seems foolish to use a data converter like
the Mohawk.
They were generally used in data entry operations as key-to-tape
rather than key punch.

or key-to-disk, key-disk-tape, key-disk-modem; MDS stuff was/(is?)
pretty hardy and pound-for-pound did what they did at a substantially
lesser cost than using the mainframe directly, especially offsite of the
mainframe.

rpl

[h***@bbs.cpcn.com]

Post by rpl
For small jobs, tapes are bulkier and require more actual
care-and-feeding than a little deck. Imagine the work required to
mount/dismount/keep-track-of 100 tapes as opposed to 100 1" thick card
decks which could just be stacked up sequentially in the hopper. By the
time you've taken the ring off the tape and opened the tape drive the
card job's already been inputted, rubber band wound and tossed back onto
the "pickup" table.

Very true. As long you have a spooler so the CPU isn't bound by
waiting on the card reader, there's no problem feeding in jobs at
the card reader. University data centers often had extra card readers
in student areas--sometimes self service, sometimes operator attended--
where students would load in their jobs. Another area would have bins
to receive printed output.

Post by rpl
or key-to-disk, key-disk-tape, key-disk-modem; MDS stuff was/(is?)
pretty hardy and pound-for-pound did what they did at a substantially
lesser cost than using the mainframe directly, especially offsite of the
mainframe.

I forgot that compared to an 029 those machines could do a lot more.
First, they weren't limited to an 80 character record. Second, they
could be programmed with formatting and editing so the keypunch
operator
is told right away of a potential error (ie in a date or code field)
rather than waiting for the batch to be processed. Third, they were
much quieter. The 129 was a long overdue improvement (due to having
buffering) but still didn't compare to the other systems.

[rpl]

Post by h***@bbs.cpcn.com

Post by rpl
For small jobs, tapes are bulkier and require more actual
care-and-feeding than a little deck. Imagine the work required to
mount/dismount/keep-track-of 100 tapes as opposed to 100 1" thick card
decks which could just be stacked up sequentially in the hopper. By the
time you've taken the ring off the tape and opened the tape drive the
card job's already been inputted, rubber band wound and tossed back onto
the "pickup" table.

Very true. As long you have a spooler so the CPU isn't bound by
waiting on the card reader, there's no problem feeding in jobs at
the card reader.

I don't think I've ever worked on a system that didn't have
reader/printer spoolers. Even so the "human spooling" component
vis-a-vis the feeder makes that a mootish point

Post by h***@bbs.cpcn.com
University data centers often had extra card readers
in student areas--sometimes self service, sometimes operator attended--
where students would load in their jobs. Another area would have bins
to receive printed output.

... or the "CS student shuffle" .... lineup for a vacant keypunch
machine, key some cards, shuffle a few paces over and join the reader
line, feed the job through, shuffle a couple paces into the printer
line, get the output, shuffle over a few paces to the keypunch line,
rinse & repeat.

Post by h***@bbs.cpcn.com

Post by rpl
or key-to-disk, key-disk-tape, key-disk-modem; MDS stuff was/(is?)
pretty hardy and pound-for-pound did what they did at a substantially
lesser cost than using the mainframe directly, especially offsite of the
mainframe.

I forgot that compared to an 029 those machines could do a lot more.

MDS had a pretty extensive end-to-end data handling lineup. I'm not
familiar with their line of cardpunch machines (before my time with them
I think).


rpl

Post by h***@bbs.cpcn.com
First, they weren't limited to an 80 character record. Second, they
could be programmed with formatting and editing so the keypunch
operator
is told right away of a potential error (ie in a date or code field)
rather than waiting for the batch to be processed. Third, they were
much quieter. The 129 was a long overdue improvement (due to having
buffering) but still didn't compare to the other systems.

[h***@bbs.cpcn.com]

Post by rpl
I don't think I've ever worked on a system that didn't have
reader/printer spoolers. Even so the "human spooling" component
vis-a-vis the feeder makes that a mootish point

Plenty of low-end S/360s didn't have spooling. Spooling required
a separate region which required more core memory.

Low-end S/360 shops didn't have any 'system programmers'--there
just wasn't very much to do. DOS was very straight forward.
Adding a spooler took a little bit more talent that a small shop
might not have readilly available. We used a product called "Power";
I don't know when that came out. DOS itself didn't offer spooling.

Before S/360 the most common computer was the 1401 and that
normally had no spooling. Indeed, the 1401 was used as a front-end
spooler to large machines like the 7090 where data was loaded to tape.

Spooling unit record I/O makes a drastic improvement in productivity,
esp on machines that have only one job running in the first place.

Post by rpl
... or the "CS student shuffle" .... lineup for a vacant keypunch
machine, key some cards, shuffle a few paces over and join the reader
line, feed the job through, shuffle a couple paces into the printer
line, get the output, shuffle over a few paces to the keypunch line,
rinse & repeat.

I guess we were lucy in college in not having those lines.
We did have to wait for our job to run and wait for the printer
operator to burst the jobs and put them in bins.

Our school gave each student in a class a budget of "funny money".
Every job cost money from that account. This forced students to
deskcheck carefully less they run their account dry. The accounts
were not generous.

This arrangement penalized commuters over resident students.
The data center had a three-tier rate system--day, evening, weekend.
Residents could run jobs at offpeak hours and save money in their
accounts.

[rpl]

Post by h***@bbs.cpcn.com

Post by rpl
I don't think I've ever worked on a system that didn't have
reader/printer spoolers. Even so the "human spooling" component
vis-a-vis the feeder makes that a mootish point

Plenty of low-end S/360s didn't have spooling. Spooling required
a separate region which required more core memory.
Low-end S/360 shops didn't have any 'system programmers'--there
just wasn't very much to do. DOS was very straight forward.
Adding a spooler took a little bit more talent that a small shop
might not have readilly available. We used a product called "Power";
I don't know when that came out.

...before the 80's, I remember that, "Power-VS". But the system it was
on definitely had both reader and print spoolers (370/148 DOS/VS-E)

Post by h***@bbs.cpcn.com
DOS itself didn't offer spooling.

did on my DOS box (c.1981); maybe that was part of the "E" in DOS/VS-E;
I didn't use POWER for anything that I recall.

Post by h***@bbs.cpcn.com
Before S/360 the most common computer was the 1401 and that
normally had no spooling. Indeed, the 1401 was used as a front-end
spooler to large machines like the 7090 where data was loaded to tape.
Spooling unit record I/O makes a drastic improvement in productivity,
esp on machines that have only one job running in the first place.

what? (this time it's an "I think you've goofed" what, not a "what" what)

Post by h***@bbs.cpcn.com
Our school gave each student in a class a budget of "funny money".
Every job cost money from that account. This forced students to
deskcheck carefully less they run their account dry. The accounts
were not generous.
This arrangement penalized commuters over resident students.
The data center had a three-tier rate system--day, evening, weekend.
Residents could run jobs at offpeak hours and save money in their
accounts.

Couldn't get away with that today; students that ran out of credits
would whine too much.

rpl

[h***@bbs.cpcn.com]

Post by rpl

Post by h***@bbs.cpcn.com
DOS itself didn't offer spooling.

did on my DOS box (c.1981); maybe that was part of the "E" in DOS/VS-E;
I didn't use POWER for anything that I recall.

DOS-VS is a different operating system than plain DOS. DOS was
for System/360, DOS-VSE was for S/370 and later machines.

Post by rpl

Post by h***@bbs.cpcn.com
Before S/360 the most common computer was the 1401 and that
normally had no spooling. Indeed, the 1401 was used as a front-end
spooler to large machines like the 7090 where data was loaded to tape.
Spooling unit record I/O makes a drastic improvement in productivity,
esp on machines that have only one job running in the first place.

what? (this time it's an "I think you've goofed" what, not a "what" what)

The S/360 was designed for multi-tasking. That is, while the
CPU was waiting for an I/O operation to occur, it would be switched
to another task and work on that.

Unit-record (punched card and printer) I/O was extremely slow.
Running such jobs on the machine would leave the wait light on
almost continuously since so few CPU cycles were required. Because
the job was dependent on the speed of the card reader and printer,
tape and disk work had to wait, too, as of course the CPU.

By spooling those jobs ahead of time, the CPU could read and print
far faster to disk than to actual unit record devices. This allowed
the jobs to go through much faster. (Remember too that while one
job was running, cards would be read in for the next one and printing
went on independently.)


In S/360 DOS, you needed a separate program "Power" and a
partician for it. Our 1401 people weren't used to the idea
and were amazed at the speed improvement.

[Nick Spalding]

Post by h***@bbs.cpcn.com

Post by rpl

Post by h***@bbs.cpcn.com
DOS itself didn't offer spooling.

did on my DOS box (c.1981); maybe that was part of the "E" in DOS/VS-E;
I didn't use POWER for anything that I recall.

DOS-VS is a different operating system than plain DOS. DOS was
for System/360, DOS-VSE was for S/370 and later machines.

Post by rpl

Post by h***@bbs.cpcn.com
Before S/360 the most common computer was the 1401 and that
normally had no spooling. Indeed, the 1401 was used as a front-end
spooler to large machines like the 7090 where data was loaded to tape.
Spooling unit record I/O makes a drastic improvement in productivity,
esp on machines that have only one job running in the first place.

what? (this time it's an "I think you've goofed" what, not a "what" what)

The S/360 was designed for multi-tasking. That is, while the
CPU was waiting for an I/O operation to occur, it would be switched
to another task and work on that.
Unit-record (punched card and printer) I/O was extremely slow.
Running such jobs on the machine would leave the wait light on
almost continuously since so few CPU cycles were required. Because
the job was dependent on the speed of the card reader and printer,
tape and disk work had to wait, too, as of course the CPU.
By spooling those jobs ahead of time, the CPU could read and print
far faster to disk than to actual unit record devices. This allowed
the jobs to go through much faster. (Remember too that while one
job was running, cards would be read in for the next one and printing
went on independently.)

The 1410/7010 could do that with a feature called processing overlap. Since
there was no interrupt mechanism the foreground program had to poll to find
out when the i/o operation had finished.

Post by h***@bbs.cpcn.com
In S/360 DOS, you needed a separate program "Power" and a
partician for it. Our 1401 people weren't used to the idea
and were amazed at the speed improvement.

[Charlie Gibbs]

Post by h***@bbs.cpcn.com
This arrangement penalized commuters over resident students.
The data center had a three-tier rate system--day, evening, weekend.
Residents could run jobs at offpeak hours and save money in their
accounts.

Our computer center allowed you to specify scheduling priority
when you submitted your job. No matter when you read in your
card deck, if you specified the bottom priority your job was
held until the wee hours (and charged the cheapest rate).

[Pascal Bourguignon]

Post by h***@bbs.cpcn.com
I wonder if that was a marketing decision by IBM so as to
keep older tab machines still in service earning rent, or
was necessary due to the limited I/O capability of electronic
computers. Or maybe the electronic computer could only do
so much work and the rest had to still be done the old way.

At that time, the CPU were expensive. All the work you could do
without the CPU meant less processing costs. It was much cheaper to
hire a couple of tabulators operators than to schedule an equivalent
job to the CPU. The CPU could have been programmed (and has been
programmed anyway later) to do the same job, but it was not a good
idea, budget-wise, at that time.

[j***@aol.com]

Post by Pascal Bourguignon

Post by h***@bbs.cpcn.com
I wonder if that was a marketing decision by IBM so as to
keep older tab machines still in service earning rent, or
was necessary due to the limited I/O capability of electronic
computers. Or maybe the electronic computer could only do
so much work and the rest had to still be done the old way.

At that time, the CPU were expensive. All the work you could do
without the CPU meant less processing costs. It was much cheaper to
hire a couple of tabulators operators than to schedule an equivalent
job to the CPU. The CPU could have been programmed (and has been
programmed anyway later) to do the same job, but it was not a good
idea, budget-wise, at that time.

And the "sceduler" was an operator who had priorities about
whose job got run first. This made operators gods who could
make your grade or break you; appropriate offerings were
submitted with all card decks :-)).

/BAH

Subtract a hundred and four for e-mail.

[rpl]

Post by j***@aol.com

Post by Pascal Bourguignon

Post by h***@bbs.cpcn.com
I wonder if that was a marketing decision by IBM so as to
keep older tab machines still in service earning rent, or
was necessary due to the limited I/O capability of electronic
computers. Or maybe the electronic computer could only do
so much work and the rest had to still be done the old way.

At that time, the CPU were expensive. All the work you could do
without the CPU meant less processing costs. It was much cheaper to
hire a couple of tabulators operators than to schedule an equivalent
job to the CPU. The CPU could have been programmed (and has been
programmed anyway later) to do the same job, but it was not a good
idea, budget-wise, at that time.

And the "sceduler" was an operator who had priorities about
whose job got run first. This made operators gods who could
make your grade or break you; appropriate offerings were
submitted with all card decks :-)).

You've got ops mixed up with keypunch; keypunch required *much* more
stroking: herbal tea for the supervisor and donuts/muffins at least once
a week; ops could be had with an annual bottle of whisky for their
manager and the occasional case of beer for the rank-and-file. Between
that and flowers for the IT Manager's secretary (to keep him out of
your hair), being a programmer was an expensive proposition; though to
be sure occasional ball-game (or up in this neck of the woods, hockey)
tickets were never out of place.


rpl

Post by j***@aol.com
/BAH
Subtract a hundred and four for e-mail.