Here is a place to discuss about Sage infrastructure, its decision making, its hosting, its policy,...

Infrastructure management workflow

What is wrong with the current workflow

  1. someone, not necessarily involved in Sage community (e.g. student, askbot developper) set up a service. S┬Ěhe choses the tool independently from the others services (e.g apache vs nginx, modcgi vs uwsgi, crontab vs rc.local vs init.d vs cron.d, mysql vs postgresql). It is undocumented. But it works.
  2. as long as it works, no maintenance is done
  3. at some point, the service crashes or get spammed
  4. William Stein sends an email to call for volunteering to fix the issue
  5. hopefully, a one-day-hero shows up, and tries to fix the issue. Because nothing is documented, a huge work of archeology is required (each time). Fixing the issue usually adds one layer of undocumented and non-uniform choices.
  6. if it succeeds, goto 2.
  7. if it does not, perhaps another one-day-hero tries to fix the issue. Sometimes, this ends-up with the end of the service and users losing some of their work (e.g. sagenb.org). Sometimes it takes months (e.g. ask.sagemath.org).

What would be a better workflow

All infrastructure administration tasks should be done by a Sage engineering task force collective, with well identified tasks, in a transparent and documented way, so that there is no dependence on some people, that puts pressure if that person is not available (e.g. they should not apology to attend a conference for some weeks!).

Some more specific proposals:

Infrastructure hosting

Advantages of academic (University or research networks) hosting

The proposal is to spread Sage services (including replication and backup) among universities worldwide, administered by a dedicated team of Sage developers (see above).

Here are some more precise benefits:

Invalid arguments against academic hosting

Advantages of corporate cloud hosting

Debate/Collective infrastructure management (last edited 2016-08-25 21:22:18 by tmonteil)