On Mon, May 18, 2015 at 9:29 AM, William Stein <wst...@gmail.com> wrote:
> On Mon, May 18, 2015 at 8:37 AM, Thierry > <sage-googlesu...@lma.metelu.net> wrote: > > Hi, > > > > On Mon, May 18, 2015 at 01:49:22AM -0700, Harald Schilly wrote: > >> > Is Sage living in a state of emergency ? > >> > >> Yes > > > > OK, so we should just breathe, give more details about the issue, call > for > > help, and work all together. The Sage community is pretty wide, and > > hosting a VM is not a big deal for a university. > > Harald also submitted a $5K grant application to Google for Sage (not > SageMath, Inc.) last week, which would pay all hosting easily for a > while. We haven't heard back yet from Google about that. We also > may use the GSoC mentor stipends (about $2K) to cover hosting for a > while. > > > > >> There were emails about this, for weeks. tl;dr: power button -> off > > > > Could you please provide more details, in particular what is far from > > clear to me is which services are affected with this shut down. I do not > > want other services like git/trac/wiki/ask/... leave our hands just > > because we did not find time to turn over. > > I'm optimistic that this won't happen. I don't necessarily have > control though, and I strongly recommend that we get organized and > move all services out of UW. > > > Having the control of our > > infrastructure is pretty important (i do not want to think of what mess > > will happen when googlegroups will be shut up by lack of income (as > > googleID did recently)). > > googleID --> openid. It was not shut down due to lack of income, but > because there is a much better replacement that has been around for > years. Google didn't want to support an old and broken auth system, > when they are also supporting a much better one. They provided a lot > of time for people to switch, and a message on every login. The issue > for sagenb at least was that -- despite repeated requests and emailing > like 20,000 people (!) -- nobody actually had the time, resources, and > expertise to take over sagenb and redo the auth system in time. That > did change though, and somebody has volunteered to do it, and I expect > to see sagenb appear again sometime this summer (when they aren't busy > teaching). > > > I started a page on the wiki about our infrastructure so that we are less > > blind, as you will see, there are quite a few points that i could not > > discover myself. > > > > http://wiki.sagemath.org/Infrastructure > > > > Please do not hesitate to fill missing and incorrect entries if you know > > more. Also, please point which service will be affected by the shutdown, > > and the ressources required to maintain them. > > > >> Representatives of universities are here? I didn't know that. > > > > There are a lot of permanent researchers/teachers from all over the world > > here, and as such, they can ask for some support to their institution. > > > > To give some recent sucessful examples: > > - Ohio State University is now hosting ask.sagemath.org > > - University of Frankfurt is now hosting sagecell.sagemath.org > > > > This is the right way. Decentralization of services among various > > universities is sustainable as they will not shut up simultaneously. > > > > Indded, we have lost ask.sagemath.org for quite a long time and public > > notebooks on sagenb.org forever because some spam on those could imply > > total shutdown on other services, this would not happend in a more > > distributed system. > > Having people volunteer to host things at random universities won't > solve the problem above in the least. You would need to also rewrite > the relevant services, e.g., ask, sagecell, sagenb, etc., to be > decentralized. This is not easy in some cases... > > It's also worth keeping in mind that people may offer to host services > at their university, unaware that what they are doing might > potentially violate rules. At least, I know from personal experience > that this could be an issue.. It's interesting to compare: > > 1. Univ of Washington's policy on use of equipment at [1] -- > > with > > 2. Google's cloud computing use policy at [2] -- basically, don't > send spam or break the law. > > Regarding 1, UW's policy says "University resources, including > facilities, computers, and equipment, may not be used for [...] for > personal benefit or gain, or for the benefit or gain of any other > individuals or outside organizations. [...] Supporting, promoting, or > soliciting for an outside organization or group unless otherwise > provided by law and University policy.[...] Assisting an election > campaign, promoting or opposing a ballot proposition or initiative, or > lobbying the state legislature.[...] Advertising and selling for > commercial purposes. [...] online book sales or referral links to book > publishers or sales houses". > > Wow, every time I look at [1] I realize there's another rule I didn't > know about that is frequently violated. For example the last rule > mentioned above seems to say that if I publish something, I can't > provide a link to the publisher. That's frustrating given that > publishers frequently allow a pdf of an article/book on my website > only if I provide a link to the publisher. UW's rules may seem > insane, but they might not be that different from other public state > universities. Here's a link [3] to a prosecution of a full professor > [4] at another Washington state university. When I talk to the > locals (e.g. other professor, my chair, etc.) about this stuff, I > expect they're going to tell me I'm being paranoid and overreacting, > but in all cases they say the opposite... that things are trickier > than I imagined and getting worse. > > I've wondered for years why so few people provide online computing > resources to the math research community... why I was the main person > to do be doing this (since I started in the 1990s). Now maybe I know > -- I was just really naive. > > The legal situation is probably much different at private universities > and universities outside of the USA. > > [1] http://www.washington.edu/admin/rules/policies/APS/47.02.html > [2] https://cloud.google.com/terms/aup > [3] > http://www.ethics.wa.gov/ENFORCEMENT/Results_of_Enforcement/2012-040%20Pezeshki%20Stip.pdf > [4] http://www.mme.wsu.edu/people/faculty/faculty.html?pezeshki > > > > >> Also, we do not need a fallback server, but a main server. > > > > Which services ? Which ressources (how many VM, bandwidth, RAM, HD for > > each) ? I just met the sysadmin of my new lab and we can host quite a few > > in terms of bandwidth and RAM. There is some short-term issue regarding > > hard disk size because of a migration currently being done, but i guess a > > few GB are OK (probably not the full history of all the Sage binaries > > ever). He is pretty competent and i had no problem to set up a VM for an > > experimental index of Sage ressources (see ##16114), there will be a > > separate network set up for Sage. > > > > So, at least in my university, we have the resources to maintain part of > > our infrasctructure (say git/trac and perhaps another one, though on the > > long term it would better for each university to host a single tool). And > > i am certain that it is the case for other universities, they have a > > mission of public service. > > How refreshing and awesome! France is awesome. :-) > > As mentioned above, we have paid-for Google Compute Engine resources > -- if somebody will volunteer to move trac and the wiki, I (or > Harald!) will spin up a VM for them to do so immediately. I can move trac during sage days next week. > Google > dropped their prices on standard instances 20% today, making this even > more feasible -- it would cost about $65/month for a VM to run trac + > wiki, which would fit in the budget paid for by GSoC... Running on > GCE has some advantages, e.g., it's trivial and very cheap to make > many snapshot backups of the underlying disk image, which are globally > distributed. We can also easily reconfigure the size of the VM based > on changing needs. > > > We should also think on a way to share cross-backups among hosts (local > > ones are good, but they got destroyed if the building burns). > > Google cloud storage is also useful -- it's $0.01/GB per month -- and > is highly redundant and globally distributed. One way I use it for > backup is to create > a local backup using bup (=git but you can store anything in it), then > rsync the bup repo to google cloud storage. I backup SageMathCloud > using a combination of about 1.5TB of Google Cloud storage (so > $15/month), plus backups to several encrypted USB drives under my > physical control, which are often not plugged into any computer at > all. No matter what backup strategy you have, I personally think it > is critical that there is a component that involves some USB disks > that are physically detached from any computer. Here's a recent > famous example of a company that kept all their backups online and > died as a result [5]. > > [5] > http://arstechnica.com/security/2014/06/aws-console-breach-leads-to-demise-of-service-with-proven-backup-plan/ > > > > > Ciao, > > Thierry > > > > > > > > -- > > You received this message because you are subscribed to the Google > Groups "sage-devel" group. > > To unsubscribe from this group and stop receiving emails from it, send > an email to sage-devel+unsubscr...@googlegroups.com. > > To post to this group, send email to sage-devel@googlegroups.com. > > Visit this group at http://groups.google.com/group/sage-devel. > > For more options, visit https://groups.google.com/d/optout. > > > > -- > William (http://wstein.org) > > -- > You received this message because you are subscribed to the Google Groups > "sage-devel" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to sage-devel+unsubscr...@googlegroups.com. > To post to this group, send email to sage-devel@googlegroups.com. > Visit this group at http://groups.google.com/group/sage-devel. > For more options, visit https://groups.google.com/d/optout. > -- Andrew -- You received this message because you are subscribed to the Google Groups "sage-devel" group. To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+unsubscr...@googlegroups.com. To post to this group, send email to sage-devel@googlegroups.com. Visit this group at http://groups.google.com/group/sage-devel. For more options, visit https://groups.google.com/d/optout.