Re: [gopher] gopher++ (gopher1) protocol
On 2010-01-11 15:02:10, Kim Holviala wrote:
> A theoretical client which purposefully splits the first two lines
> into two different TCP packets splitting them EXACTLY after the
> first CRLF gracefully falls back to standard gopher0. That's about
> as clean fallback as there can be.
You should never rely upon particular way of splitting the packages
and implemented as you write would be really wrong because I will
never know if the Random Number God will allow me to use gopher++
features today or not (depending on timing and route of packages).
That's not graceful at all. Of course in typical environment it
should work anyway but someday you may run out of luck.
> >Why not adding yet another<tab>? This way, old gopher servers won't
> >break. Just an idea.
>
> And not a bad idea at all. I just thought I could do it without
> adding anything to the selector. Unfortunately I couldn't test it so
> it's all a bit hypothetical.
I think that using additional tab and some flag after the selector would
be MUCH better in terms of robustness and ease of implementation than any
other way of signaling the extension (some timeouts or whatever). I don't
see any improvement in not doing it this way and I would strongly recommend
this.
--
Kacper Gutowski
_______________________________________________
Gopher-Project mailing list
Gopher-Project@lists.alioth.debian.org
http://lists.alioth.debian.org/mailman/listinfo/gopher-project
Reply to: