public-inbox-httpd feature request: sort oldest-first

public-inbox-httpd feature request: sort oldest-first

[Nicolás Ojeda Bär]

Hello,

Unless I missed something there does not seem to be a way to sort the
message list oldest-to-newest. A little arrow somewhere to invert the
order would be nice to have.

Thanks!

Best wishes,
Nicolás

Re: public-inbox-httpd feature request: sort oldest-first

[Eric Wong]

Nicolás Ojeda Bär <n.oje.bar@gmail.com> wrote:
> Hello,

Minor nit about the Subject: this affects the PSGI
PublicInbox::WWW module, which works best with the bundled
-httpd, but really can work with any PSGI
(Perl Server Gateway Interface) server.

> Unless I missed something there does not seem to be a way to sort the
> message list oldest-to-newest. A little arrow somewhere to invert the
> order would be nice to have.

Right; it's always favors newest messages.  We usually find
newer messages more relevant and also git packs are optimized
for newer data.

I'll take your request into consideration, but am currently
against this:

I don't want to go down the path of making the WWW interface too
complex or trying to satisfy everybody with it.  User interface
is a PERSONAL choice and I want it to remain that way.

The most important part of the WWW interface is it has a stable
URL format for downloading mboxes and Atom feeds.  These allow
and encourage users to use their favorite tools locally to
consume them.  Supporting NNTP is another key to encouraging
users to choose their own UIs.

To that end, the bundled -httpd and -nntpd servers are designed
to handle aggressive clients fairly even if they simultaneously
download huge amounts of messages.

Thanks for the interest.

Re: public-inbox-httpd feature request: sort oldest-first

[Nicolás Ojeda Bär]

On Wed, Mar 14, 2018 at 8:13 AM, Eric Wong <e@80x24.org> wrote:
> Nicolás Ojeda Bär <n.oje.bar@gmail.com> wrote:
>> Hello,
>
> Minor nit about the Subject: this affects the PSGI
> PublicInbox::WWW module, which works best with the bundled
> -httpd, but really can work with any PSGI
> (Perl Server Gateway Interface) server.

Thanks for the clarification.

>> Unless I missed something there does not seem to be a way to sort the
>> message list oldest-to-newest. A little arrow somewhere to invert the
>> order would be nice to have.
>
> Right; it's always favors newest messages.  We usually find
> newer messages more relevant and also git packs are optimized
> for newer data.
>
> I'll take your request into consideration, but am currently
> against this:
>
> I don't want to go down the path of making the WWW interface too
> complex or trying to satisfy everybody with it.  User interface
> is a PERSONAL choice and I want it to remain that way.

I share this point of view. Having said that, _personally_ I still
feel reverse date ordering is an important feature to have in the web
interface.

> The most important part of the WWW interface is it has a stable
> URL format for downloading mboxes and Atom feeds.  These allow
> and encourage users to use their favorite tools locally to
> consume them.  Supporting NNTP is another key to encouraging
> users to choose their own UIs.
>
> To that end, the bundled -httpd and -nntpd servers are designed
> to handle aggressive clients fairly even if they simultaneously
> download huge amounts of messages.
>
> Thanks for the interest.

Thanks for your work!

Nicolás

Re: public-inbox-httpd feature request: sort oldest-first

[Eric Wong]

Nicolás Ojeda Bär <n.oje.bar@gmail.com> wrote:
> On Wed, Mar 14, 2018 at 8:13 AM, Eric Wong <e@80x24.org> wrote:
> > Nicolás Ojeda Bär <n.oje.bar@gmail.com> wrote:
> >> Unless I missed something there does not seem to be a way to sort the
> >> message list oldest-to-newest. A little arrow somewhere to invert the
> >> order would be nice to have.
> >
> > Right; it's always favors newest messages.  We usually find
> > newer messages more relevant and also git packs are optimized
> > for newer data.
> >
> > I'll take your request into consideration, but am currently
> > against this:

So I've realized reversing results can be faster than "git blame"
in some cases, so I've implemented this:

  https://public-inbox.org/meta/20190615202342.7730-1-e@80x24.org/T/

> > I don't want to go down the path of making the WWW interface too
> > complex or trying to satisfy everybody with it.  User interface
> > is a PERSONAL choice and I want it to remain that way.

So I didn't state this before; but one of my concerns was about
extra parameters and inconsistent ordering leading to duplicate
cache entries (in either the client or varnish).

So I managed to overload the "o=" paramter to use negative
values, which makes sense to users of Perl (and similar)
languages which allow negative array indices to look backwards.

I'm actually regretting making "x=" a separate parameter,
instead of just replacing "q="...

So I might also start 301-ing search requests to enforce a
consistent ordering for cache-friendliness...

> I share this point of view. Having said that, _personally_ I still
> feel reverse date ordering is an important feature to have in the web
> interface.

Fair enough.  I'm considering developing some sort of REST API or
just for search; since their's no provision for it in NNTP; and
IMAP search capabilities are too limited.