Re: Next steps for rfc6874bis

[Brian E Carpenter <[e-mail]>]

Hi Michael,
On 12-Aug-21 02:28, Michael Richardson wrote:
> 
> Brian E Carpenter <[e-mail]> wrote:
>     >> Brian E Carpenter wrote on 01/08/2021 23:17:
>     >>> During the original discussion in 2012, and again recently, it was
>     >>> suggested to change the delimiter from "%" to something else.  (The
>     >>> discussion in 2012 suggested "-", e.g. http://[fe80::abcd-en1].)
>     >>> This was rejected in 2012 because it "Requires all IPv6 address
>     >>> literal parsers and generators to be updated in order to allow simple
>     >>> cut and paste; inconsistent with existing tools and practice."
>     >>>
>     >>> Do we want to revisit this?
>     >>
>     >> Definitely, but can we deal with more tractable issues first, e.g.
>     >> global warming, slaac vs dhcpv6, world peace, etc?
> 
>     bc> Since nobody has argued against Nick, can we put this issue to one
>     bc> side (as we did 9 years ago)? If we do that, we can have a fairly
>     bc> straightforward discussion with the URI and browser community,
>     bc> focusing entirely on % vs %25.
> 
> I disagree.
> I think that changing the delimiter should be on the table in the discussion.
> If the browser community doesn't like it, then fine, but let's let the group
> make the decision.

OK, but maybe we should update the draft to state that clearly as an
open issue.

> 
> My argument is:
> 
> 1) in order to maintain cut&paste, if we do %25, then we have to change
>    everything anyway.

I agree; it would be wonderful if the browser implementors could deal
with a non-encoded % but that is unclear right now.
 
> 2) people using command-line tools probably can cope with swapping around
>    stuff in the short-term.

Certainly it's better than no way of using a browser at all.

> 3) On GNU-Linux glibc systems, it's probably an update to getaddrinfo() 
only.
>    A new character is probably easier to put in than %25.

Much easier. I would never envisage supporting percent-encoding in
the socket API.

>    That could happen really quickly.

I'm not convinced. Across all o/s and all libraries? Quickly?

> Of course, it doesn't have to stop
>    understanding %.
>    Ditto BSD systems.  Few utilities parse that stuff themselves.

I agree, although in my GRASP code, I needed to parse it on Windows
until a fairly recent Python change. And in Posix systems, it has
to be done too. GRASP is very dependent on getting link-local
stuff right.
 
> 4) If we'd started in 2012, we'd already be done.  We don't have
>    interoperability now anyway.
> 
> [I abhor %25, btw]
> 
> --
> Michael Richardson <[e-mail]>   . o O ( IPv6 IøT consulting )
>            Sandelman Software Works Inc, Ottawa and Worldwide
>