[PD-dev] networking updates

[Roman Haefeli]

Hey Dan

On Sun, 2019-08-11 at 03:49 +0200, Dan Wilcox wrote:

> Christoph and I have more or less finished some work that updates
> Pd's networking and also fixes some  bugs and a couple pain points:

Cool. Thank you both!

I did test a few things with patches I just made up (on Ubuntu 18.04),
since my network-based patches usually use iemnet. Although, I can't
help you out with real-world data, here are some findings:

> * IPv6 support

I wasn't able to test yet. I was worried about the future with Pd and
networking, but don't have to be now, thanks to you ;-)

> * multicast

Works for me, between different computers in the same LAN and also
between different software (nc -> Pd, Pd <-> Pd).

> * netsend: optional from hostname & port outlet

That's really useful and absolutely necessary when dealing with
multiple incoming connections. Only now it is possible to write server-
like patches with Pd-vanilla.

> * netsend: connectionless UDP, no more connection refused (fire &
> forget without having to manually reconnect)

Although the old behavior was "correct" in some sense (it was
nitpickingly correct to terminate a link to an endpoint detected not to
be existent), I think the new behavior is much more useful. It works
well for me, I can shut down the receiving host, restart the software
or even replace the software by another listening on the same port, the
packets still arrive. 

> * netreceive: settable timeout which defaults to 10 seconds (no more
> super-long frozen Pd)

Definitely an improvement, too. The OS defaults are way too long. The
default is good, having it settable even better.

> * improved error printing with netsend and netrceive Find Last Error
> support
> * various small bug fixes (no more polling errors after socket is
> closed)

Nice.

> One current limitation is that Tcl 8.5 does not have IPv6 support, so
> the core communication with the GUI remains IPv4.

I wouldn't know why the internal connection should switch to IPv6. My
gut feeling is that such things will stay on IPv4 for quite long.


One little thing bugs me in the help file of [netsend] and I just
noticed that it was there since at least 0.47: The different modes are
labeled 'ASCII' vs. 'binary'. I think 'ASCII' is misleading in that
there are semantics in the message format. Spaces have meaning,
messages are always terminated by semi-colon. You can't use the 'ASCII'
mode for generating arbitrary messages. 'ASCII' suggests you could use
it to write valid HTTP requests which you can't. I think 'ASCII' should
be re-labeled as 'FUDI', since that it is the protocol the non-binary
mode actually uses.

I think these are _very_ valuable improvements for Pd's net objects. I
can't think of a reason not to include them. Of course, it would be
interesting to get some feedback also from people who actually have
patches that heavily use those objects.

Roman
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 488 bytes
Desc: This is a digitally signed message part
URL: <http://lists.puredata.info/pipermail/pd-dev/attachments/20191011/0935d04c/attachment.sig>