I've read:

A common misconception in socket programming is that \n eq \012 everywhere. When using protocols such as common Internet protocols, \012 and \015 are called for specifically, and the values of the logical \n and \r (carriage return) are not reliable.

before, and stuff around it, and I still don't understand :-(

Is this telling me that when Perl interpolates the string "\r\n" that it may use some other values for the two escapes ? If so, where ? I've not been able to catch it doing that, but I have only got Winders and LINUX to go on. Does "\n" interpolate to "\x0D" on a Mac ?

I understand that IO layers may translate "\x0A" bytes to and from "\x0D\x0A" pairs and vice versa. I guess there's similar support for "\x0D" line-endings (though I haven't found it). I can see that means that if you think of "\n" as interpolating to a local system newline, you may be disappointed. But that's not the same as "\n" interpolating to something different in different places or at different times.

Desperately seeking enlightenment....

A key quote from the documentation is "Perl uses \n to represent the "logical" newline" which implies that \n may not be \x0a. It then goes on to say that in MacPerl \n is \x0d.

By default Perl's text IO layers translate between the host OS's notion of a line end sequence and the character used internally to represent \n. If a string being manipulated hasn't been through an appropriate IO layer you will be dealing with the original CR and LF characters that were used. In various situations that means that an interpolated \n will not match LF (\x0a) and in many more situations it will not match an actual line end sequence (in Windows in particular).

As with many things in Perl, the default behavior is almost always what you want. So often so in fact that you tend to forget that there might be any other behavior.

---

Perl reduces RSI - it saves typing

Thank you.

So, just to hammer this home: in MacPerl the string "\n" really is "\x0D" -- and "\r" is "\x0A". This is not something the IO layers have anything to do with. A literal string containing "\r" or "\n" in MacPerl has a different value to the same string on, say, a Linux Perl. Gosh.

So, I suppose for MacPerl the IO layers must be capable of translating "\x0D\x0A" and "\x0A" to/from "\x0D" to allow for reading "foreign" files.

I confess I think it would be less confusing if "\n" meant "\x0A" at all times, and translation to/from system line-endings was relegated to the IO layers. So rather than "\n" being a "virtual" line ending (and "\r" being its dual), I think it would be clearer if the IO layer "normalised" line endings to "\x0A" AKA "\n" -- which is essentially what the DOS/Winders Perl does.

Ah well. There was something about this that didn't quite fit together, but without a Mac I could not pin down -- perhaps I couldn't see what the documentation was trying to tell me because I was refusing to believe in escape sequences with magical shape shifting properties :-(

Perl never ceases to amaze me.

The only system that's different from unix these days is Windows. The conversion is disabled by :raw which should be used on a socket in the first place. If any other odd systems need to be supported, it should be done by a PerlIO layer like it is in Windows. It's my opinion that the referenced section of perlport is outdated.

Does "\n" interpolate to "\x0D" on a Mac ?

Yes, on old MacPerl. Not anymore, though. (Not since OS X in 1999?)