Re: libpq parameter parsing problem

On Thu, Jan 16, 2020 at 4:49 AM Michael Paquier <michael(at)paquier(dot)xyz> wrote:

>
> > Links to authoritative sources (perhaps an RFC in this case) would be
> > better.
>
> No objections to your suggestion in this case though. I can see
> Section 2.1 from RFC3986 for that:
> https://tools.ietf.org/html/rfc3986#section-2.1

Yes, this makes more sense, especially since we already refer to that same
RFC earlier on the page:

> URIs generally follow <ulink url="https://tools.ietf.org/html/rfc3986">RFC
3986</ulink>,...

For a moment I thought we could improve further if we rephrase "symbols
with special meaning" as "reserved characters", which are defined in
section 2.2 of the RFC. But that set is broader than what actually needs
to be encoded for correct interpretation by our parser.

At the same time, we could still be more specific if we would say
"delimiters" instead of the generic "special meaning". Should we then
provide an exhaustive list of delimiters or is it clear enough like that?
For example, the whitespace doesn't need to be percent-encoded (it doesn't
hurt as you might be able to spare the quoting if using it as an argument
to a shell command), while the "equal sign", when used in the query string
part, does need encoding.

--
Alex