[nanomsg] Re: Port sharing, tcpmux, web-related roadmap et c.

  • From: Martin Sustrik <sustrik@xxxxxxxxxx>
  • To: nanomsg@xxxxxxxxxxxxx
  • Date: Thu, 20 Nov 2014 09:58:36 +0100

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 20/11/14 07:10, Jason E. Aten wrote:

> 1. I see RFC 1078 is generally condemned as a security risk due to
> the HELP command returning a catalog of registered servicepoints, 
> facilitating network scanning.  Can that be disabled, at least by
> default?

Yes. There's no HELP command in the implementation.

> 2. The case *in*sensitivity of the 1078 service names would seem
> to create more problems than it solves, and seems a poor choice. I
> would prefer case sensitive names and some kind of specification of
> the name strings binary format (utf8? ascii?)

The spec was written in 1988, so I guess it's ASCII. Still, RFC 1078
is far from perfect.

I've implemented it mainly because it is very simple. In the end, we
maybe want to use WebSocket multiplexer instead.

> 3. Bad experience: I've had even UUIDs collide. Hence it would be
> great if the service could provide atomic registration of a 
> known-not-to-collide name prefix or suffix, so there's no
> probability, however small, of failure. With tcpmux it should be
> trivial to implement, and provide clients iron-clad guarantee of
> that this name is a true nonce, never before and never again
> issued. This would be a strong selling point, making client side
> programming much easier.

How would you do that? IANA-style registration or something
programmatic rather?

On a kind of related topic, I was thinking about binding to whole
namespaces with WebSocket multiplexer. So, if you bind to URL
"/services" you would get all the connections to "/services/foo",
"/services/bar" et c.

> 4. Is there sample client code available?  The callback/second half
> of a service-provider code is probably here on stack overflow.  Is
> there sample code as to how would I register with tcpmux prior to
> the callback?
> 
> http://stackoverflow.com/questions/2358684/can-i-share-a-file-descriptor-to-another-process-on-linux-or-are-they-local-to-t

It's
> 
done inside of nanomsg, just use "tcpmux" transport.

Or do you want to use the multiplexer outside of nanomsg? If so, the
code is pretty trivial. I can put it online in case of need.

Martin
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.11 (GNU/Linux)

iQEcBAEBAgAGBQJUba08AAoJENTpVjxCNN9YIHQH/jUZ8d61VJObrFJwk+aP8PXW
RFD5xrnsa/ZxzvOspe/XsW9W5Y5SGdBMSwpkxNAWSb7ogyFN7svkKMkPb8vNizmH
D4F3x06ZQmDMaRDNtrSfLKzHVhGRwBW3CwZTQsDePix8d/8kEYo7FpBT6aVsZWDK
+989xwlci6gKYtoFa+guEHlzf2EoNckcKZoMnzHQQSIhl7aLpcC52nIB4aoQg9Vr
XoFn/I7l/4GDnnb5H72oN38nRtN1tFyE93RQxS/zEf51i38Zr4EeejPWexn2GOdF
F6wP1/lxI9MMUP/b1VG3AxdT73RRUnf23vr6Vd8DCtlvBTvD31kvQBzib7zuop8=
=i5CS
-----END PGP SIGNATURE-----

Other related posts:

  1. Home
  2. nanomsg
  3. Archive
  4. 11-2014