[08:30:00] Is running the LanguageTool binary (Java app) https://github.com/languagetool-org/languagetool via Shellbox something that is theoretically possible? Or is it a bad idea to run a Java app via Shellbox? [08:35:55] kostajh: without context, my understanding is that webservice stuff would ideally run on k8s, command line stuff on shellbox, but serviceops & securityops may know better [09:09:46] <_joe_> kostajh: it can be possible and a bad idea :P [09:10:22] <_joe_> kostajh: specifically, https://dev.languagetool.org/http-server [09:10:31] I thought command line stuff on shellbox = running on k8s? [09:10:34] <_joe_> so it seems like this thing has an http server builtin [09:11:19] ok a bit of context. We are building a front-end feature in MediaWiki to guide users in making "copy edit" type of edits. The backend for this thing would eventually be an application that is deployed on k8s, probably wrapping LanguageTool and providing some filtering of results [09:12:22] but because there is a desire to get a proof-of-concept together quickly, I was wondering if we could use shellbox to call aspell/hunspell or possibly the LanguageTool java binary. Then a few months later we would switch to using the yet-to-be-written service application on k8s [09:12:52] _joe_: yes, LanguageTool has both an HTTP server and a CLI, both wrapping the same Java application [09:13:11] <_joe_> so yeah I would say the right way to do it is to just run the http server in production [09:13:36] <_joe_> we build a docker image, a small chart, slap it beyond ingress [09:13:40] ok [09:13:50] <_joe_> the advantage is you don't have to change much once you're [09:13:59] <_joe_> ready to have something fuller [09:14:01] that sounds fine. [09:14:01] <_joe_> but [09:14:16] what about using shellbox with aspell/hunspell? Is that something we could do easily? [09:14:30] <_joe_> that's surely something we can do in 10 minutes [09:15:00] <_joe_> let me check with james where he is with the transition of shellbox to php 7.4 [09:15:15] <_joe_> but it's matter to add an additional flavour to the shellbox pipeline [09:15:28] <_joe_> then create a deployment in production. [09:15:34] nice [09:17:55] Ok. I am not sure we would actually want to use shellbox + aspell, but good to know it is straightforward to do. I'll know more after talking with Research folks later this week. [09:36:17] I am testing https://github.com/facebook/zstd, in case someone else is interested on how it performs in real life scenarios [09:37:40] jynus: worth noting that the licensing seems problematic with that - COPYING is GPLv2, LICENSE is BSD 2-clause [09:38:18] Hm, README says dual-licence, I'm not sure how well that is expressed legally [09:38:21] that's weird because I actually downloaded from Debian repo directly [09:40:32] https://metadata.ftp-master.debian.org/changelogs//main/libz/libzstd/libzstd_1.5.2+dfsg-1_copyright says BSD-3-Clause and GPL-2, so I guess it's OK [09:41:16] don't know much, but maybe COPYING is a standard/requirement of GPL? [09:42:27] https://github.com/facebook/zstd/pull/801 [09:42:59] <_joe_> Also I prefer to let lawyers tell me if licensing is problematic, as long as the intention to make something freely available is clear [09:43:11] <_joe_> else I'd have become a debian developer [09:43:19] <_joe_> :P [09:44:09] :) [09:46:27] just to be clear- I am not advocating to migrate anything to that, but I want to test myself the claims of throughput and compression to see how realistic they are [09:46:55] <_joe_> jynus: it's indeed interesting although it would be *really* interesting for me if browsers support it [09:48:22] I got some red flags for maturity, but first I want to check if there is some marketing on their claims [09:50:20] jynus: I'm definitely not suggesting you shouldn't do so :) [09:51:13] well, they are annoying me by not following gzip/bzip/lzip command line flags, so so far they are not helping [09:51:33] heh [09:51:57] I am checking it because that got integrated with one of our backup tools (mydumper) [09:55:28] https://bugzilla.mozilla.org/show_bug.cgi?id=1301878#c20 sounds like no browser support is expected for Firefox at least [10:06:03] jynus: I've experimented a bit with zstd, was super impressed [10:06:19] with custom dictionaries? [10:06:22] no [10:07:39] unsure if it will be a good fit for me- the focus seems to be on-the-fly memory-fit datasets decompression [10:52:57] moritzm: do you mind if I merge your SPDX headers patch on puppetmaster too? [10:53:19] please do! [10:53:37] done :) [10:57:45] thx