On Tue, Feb 11, 2020 at 1:59 AM Robin Lee Powell <
rlpowell@digitalkingdom.org> wrote:
>
>
>
> On Fri, Feb 07, 2020 at 07:58:58AM -0500, Mike S. wrote:
> > On Thu, Feb 6, 2020 at 1:29 AM Robin Lee Powell
> > <
rlpowell@digitalkingdom.org> wrote:
> >
> > > Some of you have noticed problems with dynamic content on
> > >
lojban.org now that we've started using CloudFlare. I haven't
> > > been able to figure out how to fix this, so here's what's going
> > > on; maybe somebody else will have ideas.
> > >
> > > So the goal of CloudFlare, primarily, was to make it so that if
> > > my server went down, the site would be basically fine. This we
> > > have acheived.
> > >
> >
> > Judging from what I saw during the last outage, I believe
> > Cloudflare is caching only pages that have been requested at least
> > once. Assuming I am right about that, I don't know if there is a
> > way to make Cloudflare cache the whole site without clicking
> > through every page on the Wiki. By the way, I don't care about
> > this issue since the site is up most of the time anyway, but since
> > the whole point of the exercise is to cache pages and that doesn't
> > necessarily happen, I thought I'd mention it!
>
> Yeah, you're absolutely right about that. For me the point is to
> make it so that the *basics* are up no matter what.
>
As a side note, it dawns on me one could run a website-ripping tool to hit all the pages.
> The specific complaints that I was receiving were about that the
> site being down *entirely* sends a message to people unfamiliar with
> the projec that it's dead. Which is not unreasonable: if I'd just
> heard about a super-nerdy language and I went to the website and it
> was down, I would in fact assume that the project was dead,
> personally.
>
> > At any rate, your bypass rule for talk pages seems to be working okay.
> > Since we don't know what the problem is, I suggest the following for a temp
> > fix -- Please add the following rules in the proper location:
> >
> > *
lojban.org/*LFK* Cache Level: Bypass
> > *
lojban.org/*[bypasscache]* Cache Level: Bypass
> >
> > The first line will help the new committee work on their pages. The second
> > line will allow a person to work on an arbitrary page without the cache.
> > When the page is finished, it can then be moved to the correct location.
> >
> > I *believe* these rules will be sufficient for the time being and get you
> > off the hook. Your help is very much appreciated!.
>
> I'm actually wondering if we might want to go the other way:
> explicitely cache the very static stuff (_javascript_, css, etc) and
> then list specific pages that we don't mind caching long-term, like
> the front page.
>
I was sorta wondering why it was done the way it was done. Are you using CloudFlare partly to minimize the load on your server?
Also, does adding a "Bypass" rule for a page deactivate the cache entirely for that page? (That would obviously be very bad for some pages.)
If the answer to both questions is no, why not bypass all dynamically generated pages like so:
*
lojban.org/papri/* Cache Level: Bypass
*
lojban.org/*php* Cache Level: Bypass
*
lojban.org/* Browser Cache TTL: 30 minutes, Always Online: On, Cache Level: Cache Everything, Edge Cache TTL: 2 hours, Origin Cache Control: Off
... assuming, again, the caching mechanism continues to work, which we could test in order to be sure.
The first line makes the whole Wiki live when accessed through the short URL. The second any pages directly invoking php, which includes page edits (which need to be bypassed).
^Just a thought. The LFK bypass rule is working (I know because I tested it) which makes the situation a lot better. Thanks.