Thanks, Cris. I applied for an an Open Source Account via the link you provided. We'll see what happens.
Norm On Monday, June 14, 2021 at 9:31:02 PM UTC-4 Cris Perdue wrote: > For what it's worth, I see that Netlify has an official form for > requesting an "Open Source" account, > at: https://opensource-form.netlify.com/. > > Note: I write this without prejudice against other options, e.g. > DigitalOcean with free Cloudflare. It's just that I have no experience > with them. > > It asks for the following information: > > Project Name > Associated email > Open source license used > Link to code of conduct > Link to our service on your main page > Is there anything else you would like us to know? > > I would expect that the account would be free, or at least free up to some > limits, if approved, but that is only a guess. > > If I were requesting such an account, I would want them to know that the > total size of files in the site is (apparently) somewhat over 1GB and that > it sometimes needs more than 100GB transfer per month. They don't list > "public domain" as a license, so that would need a mention, and surely > something about the subject matter and goals of the project, extremely > brief and untechnical; also something about the longstanding nature of the > project. > > "Code of conduct" probably translates to Metamath info for contributors. > > (One way to check disk usage would be to do "du -sh" at the shell prompt > at the root of the website directory tree.) > > > On Mon, Jun 14, 2021 at 4:03 PM Norman Megill <[email protected]> wrote: > >> On Saturday, June 12, 2021 at 11:57:47 AM UTC-4 David A. Wheeler wrote: >> >>> All: >>> >>> Does anyone have recommendations for a better hosting site for the main >>> site us.metamath.org and/or for the build/staging site us2.metamath.org? >>> Personal experience (or close to it) preferred. Alternatively, warnings >>> about “don’t do X” are also welcome. There are lots of hosting services, >>> which makes selection more complicated :-). >>> >>> Below is the background & requirements as I know them. In the end Norm >>> decides, my goal is to give him a hand. >>> >>> --- David A. Wheeler >>> >>> =========== >>> >>> The “main” us.metamath.org site is currently hosted by site5.com. I’ve >>> learned from Norm that his current term with them expires soon on >>> 7/29/2021. Now would be a good time to plan a change, because the current >>> hosting service has serious limitations. >>> >> >> site5.com is also increasing the price by 20% starting 7/29, which is >> another reason not to renew it. This pushes it to about $10/mo (if I pay >> for 3 years in advance), up from $4/mo a few years ago. >> >> In the past, us.metamath.org reliably used only volunteer mirrors for >> many years until 2008, when I felt a little uncomfortable that I had no >> backup if they all disappeared. At that time I couldn't use my home server >> as a backup because the ISP blocked port 80 without an expensive "business >> account". So site5.com was basically a cheaper way to get port 80. >> >> My current home ISP doesn't block port 80, so my house (us2.metamath.org) >> is now also available as a backup, and the original reason for site5.com >> is no longer there. >> >> Currently there are two volunteer mirrors, at.metamath.org and >> cn.metamath.org. (de.metamath.org currently points to at.metamath.org >> for historical reasons; I may remove it.) Both have been very reliable for >> the past few years, and I've used both of them for us.metamath.org when >> the need arose. >> >> us.metamath.org requires about 10GB disk and maybe 100 to 200GB/mo >> transfers. I think 1TB/mo would provide a comfortable margin. Here are some >> services people have mentioned: >> >> Linode - 25GB disk space, 1TB/mo transfer for $5/mo >> >> digitalocean - 25GB disk, 1TB/mo transfer for $5/mo. The associated >> serverpilot.io seems to be needed for https (not clear to me) for >> another $5/mo >> >> netify.com - the free plan provides 100GB/mo transfer and $20/mo each >> extra 100GB/mo, which is a problem. The next higher plan is $19/mo with >> 400GB/mo transfer. I couldn't find their disk space allowance. >> >> I may set up a Linode account for us.metamath.org and see how it goes. >> >> >>> The current us.metamath.org hosting service doesn’t allow ssh or rsync, >>> and it doesn’t make it easy to support TLS/HTTPS (it *might* be possible >>> but it’s not trivial). Those would be highly desired. For example, ideally >>> we could just enable Let’s Encrypt and have TLS/HTTPS support. >>> >> >> If us.metamath.org is changed to use https, it means that it can't be >> switched to volunteer mirrors as a backup unless they also agree to set up >> https and maintain certificates. I have little https experience and don't >> know what it involves. >> >> >>> >>> On the good side, the current hosting service doesn’t charge extra fees >>> for extra network transfers. That’s important; everyone once in a while the >>> site gets busy. Wherever it goes to should allow transfers without just >>> shutting us off at N bytes. >>> >>> It might be great to enable Cloudflare’s free tier, or some other >>> inexpensive CDN service. That would make transfers faster & cause less >>> load. We’d need a service that’d work with that. >>> >>> I don’t think GitHub hosting would work. It’s a static site, but we have >>> too much stuff :-). There are over 100K web pages generated from the *.mm >>> files. I don’t have the numbers handy, but I believe the total site size >>> exceeds the GitHub Pages maximum of 1GB >>> https://docs.github.com/en/pages/getting-started-with-github-pages/about-github-pages >>> >>> ============= >>> >>> The “build/staging” site us2.metamath.org is where the static files are >>> currently rebuilt. That is currently at Norm’s house running Debian. >>> There’s no rush to change it. However, eventually I’d like to see that >>> somewhere outside a house, so that things can continue smoothly if >>> something happens to his house or him. The site build on us2 takes around 4 >>> to 5 hours, though I suspect that can be sped up without too much effort. >>> >> >> Well, I don't think the code itself can be sped up much easily, almost >> certainly not. say, an order of magnitude. It averages about 100ms per >> page (on a 10-year-old i5-2500 CPU), which I don't see as unreasonable. I >> am aware of some bottlenecks generating very large proofs that I plan to >> address eventually, but even with that, the overall time might only >> decrease 10% or 20%. >> >> It can be sped up several times by splitting the job onto multiple cores >> if we really need to do that. >> >> It can also be sped up a factor of 2 if we eliminate the gif image >> versions of the pages. I've brought this up in the past, and someone said >> they found the gif pages useful because copy/paste produces the actual >> set.mm ascii expressions. So I left them in because there's essentially >> no extra cost in doing so. >> >> Of course speedup is always welcome, but I don't see an urgent need to >> speed up the site build, as long as it completes in say under 24 hrs. For >> urgent special cases, I have a script that will add specific new or >> modified proofs more or less instantly. >> >> If we move us2 to a paid server to do site builds, it would need around >> 100GB of disk space (around 60-70GB currently but allowing for growth), >> which is much more than the inexpensive plans provide. It's hard to compete >> with "free" (at my house). :) >> >> BTW the site build is essentially done with the script "install.sh" >> provided with the Metamath download and can be used by anyone to recreate >> the site on their local machine. The actual build script, called >> "build-metamath-site", is heavily customized for the us2 server; it calls >> "install.sh" and also creates buffer and backup directories, so that the >> site gets completely built before it is exposed on the live site, and so >> that I can instantly switch to a previous site version if something went >> wrong. >> >> BTW2 David has the passwords to all accounts (us, us2, DNS) if something >> happens to me. >> >> Norm >> >> -- >> > You received this message because you are subscribed to the Google Groups >> "Metamath" group. >> To unsubscribe from this group and stop receiving emails from it, send an >> email to [email protected]. >> > To view this discussion on the web visit >> https://groups.google.com/d/msgid/metamath/537f32d7-ce74-4151-8992-6bb9c0018d59n%40googlegroups.com >> >> <https://groups.google.com/d/msgid/metamath/537f32d7-ce74-4151-8992-6bb9c0018d59n%40googlegroups.com?utm_medium=email&utm_source=footer> >> . >> > -- You received this message because you are subscribed to the Google Groups "Metamath" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/metamath/8b8ac3d5-b9a7-4882-9b77-e13da99ac67dn%40googlegroups.com.
