I looked again at the RS232 spec and you are correct, a 3 volt swing would be enough.
No, not right - transmitter needs at least +5V swing. A normal TTL output can't get there.
Update: All hardware interfaces have a defined "margin of error". For me to send a "true" value there is a defined voltage level that must be achieved. There is a much lower voltage level that is required for you to "receive that true value". That is the "margin for transmission error". It is completely possible for me to send something to you that "by the books" is not a "true" level for me to transmit, but that you will accept as such.

This is meaningless to the software folks here..strike.. If these wimpy guys could use an open collector output to get to the full 5V, then it would be fine. But then of course they could just tie that to the 12V rail also. And this would be able to source significant current at the logic "high" level. A standard TTL gate cannot source any significant current at a logic "high" level.

This whole RS232 I/F stuff is fraught with "how its done" vs "how its spec'd" stuff. We could probably have a many page discussion about this...

I'm saying that sometimes Box A and B who claim to be compatible, will not "talk". And that does happen. Last time for me was about 18 months ago - rather than buy a serial card and deal with the software hassles with that - the client elected to just buy a different computer. This "bad boy" RS232 computer just has a different job on the network.

My point is that sometimes some I/F's won't talk to one another. I probably shouldn't have pointed this out, but I did. Its not the "normal" thing, but it can happen in practice.

UPDATE: Geez what a messy looking post! Sorry about that. The bottom line is that sometimes 2 devices cannot communicate because their hardware just doesn't meet specs, even though they might yet be able to communicate with some other device. This is rare, but it does happen. That was the point.