[bug #60666] [grotty] add OSC 8 support for hyperlinks

[G. Branden Robinson]

Follow-up Comment #4, bug #60666 (project groff):

Sorry, forgot to preview--that thing was riddled with typos.

Second try.

My design and implementation thoughts.

1. As suggested earlier, I think we want two device control keywords.  One,
"osc8" to enable/disable this feature.  The other, to actually write the link
to the output stream, I think we should call something more obvious, like
"link".  Advantages of this include (a) not having to expose yet another
environment variable for this and (b) making the device control for writing a
hyperlink independent of its representation in the output.  Consider the
possibility that OSC 8 is deprecated at some point in the future in favor of
some alternative mechanism.  It will be misleading and confusing to tell the
device 'X tty: osc8 id=foo url=bar' if the output sent to the terminal doesn't
use OSC 8.  It will not be disruptive to change some per-device tmac files to
switch out 'tty: osc8' for the new flavor.  For us, this change would be
confined to tty.tmac, and people could get ahead of us in troffrc-end if they
wanted to.

2. I'd cause SGR disablement (whether by -c option, "sgr" device command, or
GROFF_NO_SGR environment variable) to also implicitly disable OSC 8; while
they aren't _fundamentally_ coupled, they do happen to both be ISO
6429/ECMA-48 escape sequences.  We don't have any way at all to support
hyperlinks in the legacy output format.  If the many fans of backspaces in BSD
world can come up with a statistically-unlikely sequence of backslashes and
underscores with which to embed hyperlinks, then we can revisit this question,
but I think it's unlikely to arise.  Moreover, if people share Ingo's concern
that SGR itself constitutes a security vulnerability, then I can't imagine
that URL embedding wouldn't arouse even more agitation.  So I'd lose the last
chunk of the diff entirely.

3. While I personally prefer the "char const" form of declaration over "const
char" because it is more regular, it is not idiomatic in the groff source tree
and Ralph Corderoy's and my suggestion to migrate was ill met on the groff
mailing list about three years ago[1][2].

Miscellaneous factors include:

90. I think it's habit that keeps people using the 7-bit forms for ISO
6429/ECMA-48.  CSI and OSC are both well-defined C1 controls and have been for
decades.  I'd kind of like to make grotty(1) emit them.  This will probably
turn up a lot of terminal emulators that don't support them. :-/  Maybe an
"8bit" device control command is also in order.  But this is out of scope.

Comments and disputation welcome as always.

Regards,
Branden

[1] https://lists.gnu.org/archive/html/groff/2018-05/msg00028.html
[2] https://lists.gnu.org/archive/html/groff/2018-05/msg00042.html

    _______________________________________________________

Reply to this item at:

  <https://savannah.gnu.org/bugs/?60666>

_______________________________________________
  Message sent via Savannah
  https://savannah.gnu.org/