[12:58:58] jbond42: so it looks like RIPE Atlas will only do HTTP probes directed at their own anchors, do you have any idea if that's something that's likely to change sometime? [13:00:03] cdanis: i thought they could do http probes *from* anchors. i.e. you can set up a http measurment but it will only every get scheduled on an anchor and never a probe [13:03:52] i dont see them allowing http measurments on the probs, it is of course possible (we had one for http.l.root-servers.org which ran everywhere) but yuo would need to build a good case and as wikipedia.org is often a filtered site in some economies i dont think they would allow it [13:04:45] sure, that makes sense [13:05:11] but even if I ask for an HTTP measurement to be run against en.wikipedia.org from one of our own anchors I get a HTTP 400 with {"error":{"detail":"There was a problem with your request","status":400,"title":"Bad Request","code":102,"errors":[{"source":{"pointer":"/definitions/0/target"},"detail":"Only anchors may be targeted"}]}} [13:05:24] yes it looks like you are right https://labs.ripe.net/Members/kistel/http-measurements-with-ripe-atlas [13:05:57] >We're also considering adding HTTP measurements to the list of "anchoring measurements" (i.e. to the set of measurements we automatically execute towards anchors) with a relatively low frequency. [13:05:59] it looks like they did that [13:06:29] might be worth asking, at least for one off mesurements [13:07:43] when I get around to setting up atlas_exporter I'll also have it export the results of the HTTP anchoring measurements towards our anchors [13:09:24] ahh, for a while i still had admin privs but they have removed it now :(. anyway they do allow ssl measurments to anything which would still get us a lot [13:09:50] yeah, will definitely add some sslcert measurements as well [13:13:46] we could post to the mat-wg list. wikimedia is a not-for-profit and is a p[osative entity. i think we could make a case but the censoring concers are likley to win in the end [13:14:30] i blan to have a word weith people f2f when i see tyhem but with moving etc i didn't managed to schedual the ripe meeting this time round [13:14:33] yup, just signed up for mat-wg, I think I'll make a post once I also have something to say about the data we already have [13:15:11] in the past I remember concerns about a) the network being abused as a DDoS vector b) the network being used for censorship research, making it dangerous for probe hosts [13:16:08] we've worked with RIPE in the past (co-authored a blog post), so we could make a pitch for a collaboration too [13:17:55] i think we could make a stong case to have wikimedia included in the core measurments especially for ping, traceroute, ssl. which means the measurments run on all probes (i think thats what you did withthe experiment right paravoid?). http though im still unsure [13:18:20] we already have anchoring measurements that are pretty close to all that [13:18:42] so maybe it's not too big a deal [13:18:44] if we did that, we could expect lots of random traffic - do we want that? [13:19:24] traceroute, in particular tcp port 443 traceroute, is something that could be handy to have history of [13:19:42] i thought the ancoring measurments only ran from the anchors i.e. about 150-300 probes vs however many thousands they have now [13:20:15] it looks like it's anchors + a lot of probes [13:22:12] https://atlas.ripe.net/measurements/2841682/#!general [13:27:50] oh ok 7k is pretty good covrage, i checthe http.l.root-servers.org and that had all probes which is ~20k. however i suspect the 7k number is applying some common flags like stable, has_ipv4, has_ipv6 etc to make sure you are not getting propbes known tobe bad [13:30:26] ahh makes sense