Re: [arch-d] FYI: closure of the IAB Stack Evolution program

[Chris Morrow <[e-mail]>]

On Fri, 23 Aug 2019 20:28:43 +0000,
Ted Hardie <[e-mail]> wrote:
> 
> [1  <text/plain; UTF-8 (7bit)>]
> Howdy,
> 
> On Fri, Aug 23, 2019 at 11:48 AM Chris Morrow <[e-mail]>
> wrote:
> 
> >
> > I had figured quic was really an admission that: "Adding a new tcp/udp
> > like protocol was much harder to swallow for the internet than just
> > re-using an existing thing (UDP)."
> >
> > quic COULD have just as easily been done over tcp, esp, ah, icmp ...
> > all of those come with some baggage though, which UDP didn't have.
> > (admittedly udp has it's own baggage)
> >
> 
> Changing TCP behavior on most systems requires modifying the kernel.

sure, and permitting <new protocol number> though/around/over/etc all
the various 'firewall' things (not to mention that security NAT
device) is also .. fun.

> Having something built so that it could be modified in user space really
> was important.  UDP is just a demux for QUIC and that's a success because
> it enabled all of the other functionality to go up into the user space.
>

agree, just saying that there are lots of reasons why: "picked udp, not X".
but that really ... as fernando said any other number would have also worked...

except that UDP has pretty universal access end-to-end without
molestation AND it's not something kernels generally do much with
except: "here ya go socket listener!"

> Just my perspective,

thanks! (I also have a perspective! which I hope gets better with new
glasses) :)

-chris