Early yesterday morning our forum software was updated. It took a while before some of our users noticed that the site was no longer usable with the stock browser on Sailfish OS.
As an immediate remedy we have reverted the site to a backup which was taken before the update took place. As an unfortunate result the comments and posts that were made after the backup have been lost. So, if you have written any posts or comments after 3:40 UTC yesterday, please consider posting those again.
why on EARTH was DELETING posts+comments for at least one day (and maybe as much as FOUR DAYS) deemed a smaller inconvenience than temporarily breaking the website on a browser that is just barely usable as it is?
especially without a backup of the deleted posts+comments that will eventually be merged back in.
i could also ask:
why would jolla devs not test on sailfish-browser before a major forum upgrade?
why would you not fully snapshot the forums immediately before a major forum upgrade?
why is there not a GIANT NOTICE about this on the home page?
deleting data from underneath users is just about the worst thing devs can do.
Is there anybody using the sailfish stock browser at all, especially for this forum? There is an excellent and native client on Openrepos. I bet barefoot anybody would have noticed the database upgrade anyways, plus, it’s would have been a temporary incompatibility as the next release will become public shortly anyways(hopefully). I really don’t think the rollback would have been necessary, but anyways,i don’t mind either:)
Just saw there are plenty of comments lost especially in the VoLTE thread. Please don’t forget to merge them in if it’s not to much work, as there was quite some activity. Thanks.
the forum managers told YOU to repost YOUR OWN comments, after THEY deleted them, in order to fix, of all things, sailfish-browser.
completely bonkers. the only thing i can imagine is that they THOUGHT they were reverting just a small layout update, and ACCIDENTALLY reverted the entire forum DB, and are now pretending it was on purpose.
by ‘unfortunate’ did you mean ‘accidental’? if so, then you should clarify, because it SOUNDS like the decision to revert the forum DB after discovering sf-browser was broken was deliberate.
if it was accidental, i apologize for making a fuss about an honest (painful) mistake.
if it was deliberate, y’all are just, like, bad at decisions.
EDIT: my comment here is more aggressive and personally offensive than i intended it to be. i was frustrated at losing content, and was lashing out. i know that this entire situation was not intended, and i apologize for my rudeness.
To me (with dev & some devops experience) it sounds like: “This cms / system is contained in the database. It is not realistically possible to revert the update without reverting the entire db.” And yes, I know about the ‘bad decision to have the system in the db’ discussion.
They have reverted the forum to a previous version of the discourser software which still works with the latest stock browser in 4.3 and earlier ones. However, that has resulted in the loss of a lot of posts made in the days between when the discourser upgrade happened and when it was later reverted. Apparently it was not possible to take a backup of the forum before it was reverted to an earlier state, hence the lost posts.
@teleshoes, @direc85, @Steve_Everett, please mind that Jolla does not run this Discourse instance, it is a SaaS provided by Discourse Inc., as noted before. As outlined there, they sell zero administration and promise zero moderation, which I assume to be major reasons why Jolla has chosen this SaaS offer.
This answers all “i could also ask:”-questions by @teleshoes: Because Jolla does not administer this Discourse instance and they are paying money in order to not pay any attention.
This also explains why fiddling with the database of this Discourse instance (taking backups, merging entries etc.) is out of scope: Jolla does not have the know-how and resources to do that, plus the contract with Discourse Inc. presumably does not cover that. What obviously was covered was a roll-back to a snapshot of the whole container (in which this Discourse instance runs) taken some time before the update was deployed.
So far this is the sad, modern world of SaaS and outsourcing in general: Companies make themselves helpless, dependent slaves of service providers, which solely fulfil their contractual obligations. And when the internal know-how is gone, it is very hard to build it up again.
But for Jolla’s deliberate decision to let Discourse Inc. roll back to a snapshot many ten hours old, just to keep the facade that “everything is fine”, which includes erasing the thread which documented the issues with the Jolla browser (being redirected to a “browser too old”-page) among a couple of other threads and many messages in extant threads, IMO @teleshoes has chosen the right words:
why on EARTH was DELETING posts+comments for at least one day (and maybe as much as FOUR DAYS) deemed a smaller inconvenience than temporarily breaking the website on a browser that is just barely usable as it is?
if it was deliberate, y’all are just, like, bad at decisions.
I have the impression that patching the browser detection to still accept the current Jolla Browser was not seriously evaluated, maybe because it would have cost extra money.
But stating this in the way quoted below feels like a slap in the face and a bad joke for someone (e.g., me) who contributed to multiple threads, which do not exist any more, plus I “missed” to keep backups of all the posts in all threads at FSO (should have learnt from the last time I was confronted with that notion) :
P.S.: IMO it is better to omit “USA style” (i.e., meaningless) apologies in some situations. After taking actions like this, they feel like an extra kick in the buttocks.
SaaS is very common; it isn’t a good or bad thing in itself. The use case must be considered, and I’m quite fine with the forum not being self-hosted at Jolla HQ This is a public forum, after all. If there is some soft of fully automatic flagging system, I’m not a fan of it at all… Discussions on this forum have been very much on topic after all AFAIK. TJC reacting to the third flag sounds quite good approach, as the reaction is “automatic” but only after users’ actions, not autonomously.
The rollback could explain why the backup was lost, too, if the backup was on the same server… Need I say more?
I hope Jolla updates the browser engine again soon, so the forum update won’t have to be postponed too long!
Edit: Managing forum isn’t the core business of Jolla after all, so some amount of hiccups is both inevitable and acceptable. Losing four days of posts is knocking the limit, however… Let’s hope the next update goes smoother
If it is not possible to separate the data from the code because it is all mixed up together in one ‘container’ (whatever that is) then this does not sound like a very good decision by the people who designed Discourser or how its implemented. This presumably means that you could never revert the code, or perhaps even a patch to the code, without reverting the data as well, and maybe vice versa?
The forum software is called “Discourse” by “Civilized Discourse Construction Kit, Inc.”: https://www.discourse.org/
I.e., there is no concluding “r”.
If it is not possible to separate the data from the code …
There sure is data- / code-separation, technically. But likely not contractually, as pointed out. Just an often occurring consequence of using SaaS.
You might take a look at the source code of Discourse at GitHub, to see how it is designed in detail.
Besides being offered commercially as SaaS, it is also Free Software (GPLv2) so anyone can run her / his / its own instance.
yea, sorry. my ‘one-to-four days’ guess was just a guess at the time.
i was basing it on the VoLTE thread. at least 10 posts over at least 26 hours were lost there, and the most-recent remaining post was four days ago, so i knew it had to be “more than 1 day, and maybe as much as 4 days”.
: