A few times now (3 or 4, in the last 6 months, I'd say), our rich text document has reverted to an old version by a day or two. This document is routinely accessed by 12 people we trust, and as far as we can tell, nobody is doing this on purpose e.g. restoring a previous version from history (a difficult thing to do by mistake, I believe).

If it's useful to know, we are able to restore the lost data from history (sometimes), and multiple people's updates were lost, not just one person's updates.

Any ideas about how this could happen by mistake or some way to prevent it?

One hypothesis I had was about merge conflicts - if somebody has an old version open and they make a change before loading the latest updates from others, there could be some edge case of the merge conflict resolution that can cause large amounts of data to be lost?

Not sure if it would be a browser extension. It's hard to tell who made the changes looking at the history, but the last time it happened we checked with the person who it appeared may have triggered the issue and the only browser extension they had something to do with changing fonts.

I don't have too many guesses, not being familiar with the architecture of CryptPad / OnlyOffice. Thanks in advance!

14 days later

Hello,

We would indeed invite you and your teammates to check the Common Issues section of our documentation FAQ first. This is indeed the kind of problem that could be caused client-side during synchronisation.

Are some of your team members using a virtual private network (VPN) ? Do maybe one has a very low quality network access?

Okay good to know - thanks! Yes, some of us use VPNs and/or try to update the doc from remote places with low quality networks.

22 days later

@alexsapps , did it happen again?

This is an interesting and important topic, so I'd like to know (and perhaps other too) how it concluded - if it concluded.

If it happended again, please give an idea, how much was reverted. You gave a time span, but how much text? :-) Also, did you use the rich text app or the (beta) document app?

Thanks.

    Hey AlexQ it happened probably 3 times total before we switched to another system. This happened in Rich Text documents. The amount of text would be in the range of 10 - 40 English sentences in various places at the top of our document, most of it in bullet point lists, many of those lists occurring inside tables.

      @alexsapps
      Addition after having read your other thread about slow history:
      https://forum.cryptpad.org/d/326-slow-to-access-document-version-history/

      In my personal experience CrpytPad gets a bit unstable the more people edit a document simultaneously (LAN access, no intenret, no VPN). :-/ I don't know if same is true for non-simultaneously edits.

      So additonal question and hint: simultaneously or non-simultaneously edits?

      And if you do tests to narrowing the root cause down: Try a test document with 3 and later with 6 people, instead of 12. Does this change something (fewer/no losses? quicker history?). Of course you would have to add more test text, since 3/6 people write less than 12 people to exclude document size as a cause and to make your test case and your 12 poeple daily todo list comparable.

        alexsapps

        Please, so I can understand this:
        a) So no more errors sicne switching to another system???
        b) What is the difference? More RAM? Faster CPU? Other/newer OS?

          AlexQ It's mostly not simultaneous. I would be surprised to see more than two people typing at the same time.

          I don't think my team will have much willingness to run such tests unfortunately. I wanted to at least figure out if it was one person's browser / network setup causing the issue but they've moved on and aren't excited about technology generally.

          AlexQ We switched to Proton Docs and now we have a new issue which is less disruptive, and that is sometimes the browser app just stops saving changes (for that person). Usually you type something and it says saving... and then it says saved. But in rare cases which I haven't figured out yet, it just wont even say saving... and just looks like it's instantaneously already saved but it's not. It happened once naturally and I reproduced it once but couldnt reproduce again. Anyway now it's just one person's updates are missing instead of everyone's between a certain time range.

          @alexsapps Sorry to hear this. :-((
          It's often difficult to get users to try open source or some less known tools.

          Personally I would exclude browser as cause (except browser plugins!). It could be network, maybe.

          If your users already moved on please "close" your CryptPad threads as a last action (final notice for the CryptPad team).

          If you are looking for an alternative:
          Etherpad worked well (fast history and no loss), but of course it's not encrypted (!).

          Godo luck.

          Thanks for your help!

          I don't see any option to close this forum thread, at least from the mobile site. Searched for "cryptpad forum close thread" and didn't find much.

            alexsapps

            My bad. What I meant:
            Moderators can mark threads, if they know that they are e.g. solved.
            So your message here is enough :-)
            (I was / am unsure, if your other thread regarding slow history should be kept active.)

            Thanks for your time and answers and again good luck with the team plan document. :-)