Your gut feeling about auto remove?
![](https://secure.gravatar.com/avatar/148a51ecc80531586e1f5530d64b988a.jpg?s=120&d=mm&r=g)
Hello, Please take this as a user survey and answer out of your gut ignoring technical details or wording and translation issues. Looking onto the "Auto-remove" tab you do see this option. "Keep one snapshot per week for the last N weeks" Does this take the current week into account even if it is not yet complete (e.g. today it is Thursday)? Regards, Christian
![](https://secure.gravatar.com/avatar/cdfdc7382d7a988a6574e78e3bb6310c.jpg?s=120&d=mm&r=g)
Hi Christian, When I read the option, what I think is "I'll have N weekly copies at all times". Of course I'm starting fresh, I know I'll need N weeks to collect N weekly copies. Looking to my BiT timeline, and compare to what my config says, the both have no resemblence to each other, however. :) What I expected was "Today + X Daily Copies + Y weekly copies after that + Z Monthly copies after that". I think the logic works a bit different currently (at least for 1.5.2). Cheers, Hakan On 11/26/24 1:56 PM, c.buhtz@posteo.jp wrote:
Hello, Please take this as a user survey and answer out of your gut ignoring technical details or wording and translation issues.
Looking onto the "Auto-remove" tab you do see this option.
"Keep one snapshot per week for the last N weeks"
Does this take the current week into account even if it is not yet complete (e.g. today it is Thursday)?
Regards, Christian _______________________________________________ Bit-dev mailing list -- bit-dev@python.org To unsubscribe send an email to bit-dev-leave@python.org https://mail.python.org/mailman3/lists/bit-dev.python.org/ Member address: hakan@bayindir.org
![](https://secure.gravatar.com/avatar/148a51ecc80531586e1f5530d64b988a.jpg?s=120&d=mm&r=g)
Hi Hakan, thank you for your thoughts. Am 26.11.2024 12:26 schrieb Hakan Bayındır:
"I'll have N weekly copies at all times"
OK, but is the current week a week no matter if it is complete or not? ;) Let's imagine your setting says 3 weeks to keep. Today we do have week number 48 in this year. What do you think? Will BIT keep the 3 backups for week 45, 46 and 47 or 46, 47 and 48? I would like to know what you expect as a user. I now what BIT is doing or try (because of bugs) to does in this case. Christian
![](https://secure.gravatar.com/avatar/bf4c44c0e1f09bfe4841d474e888facd.jpg?s=120&d=mm&r=g)
On Tuesday, November 26th, 2024 at 10:23 AM, c.buhtz@posteo.jp <c.buhtz@posteo.jp> wrote:
Let's imagine your setting says 3 weeks to keep. Today we do have week number 48 in this year. What do you think? Will BIT keep the 3 backups for week 45, 46 and 47 or 46, 47 and 48?
Personally, I would expect that there are backups for weeks 46, 47 and 48, *if and only if* the backup for week 48 was already made (i.e. we are past the scheduled point in time for the weekly backup). But I want to back up a bit: On Tuesday, November 26th, 2024 at 5:56 AM, c.buhtz@posteo.jp <c.buhtz@posteo.jp> wrote:
Does this take the current week into account even if it is not yet complete
I think this has the same problem with wording that the existing text has. I think "take into account" could mean more than one thing. Here is some more clear wording: "Keep up to X weekly backups, made at <scheduled day of week/time> each week". It's very clear from that, how it should work. But if you wanted it to work differently, then it might be harder to explain clearly :) Karl Knechtel {:>
![](https://secure.gravatar.com/avatar/c58d851d859f0fd6f9248dde793f8ad9.jpg?s=120&d=mm&r=g)
I think it should exclude the current week. That is what I would most expect just from the language. Although it is ambiguous, and you could logically be referring to the "last of the weeks that have backups", I would more expect "the last week" to be different from "the current week". If it surprises someone one way or the other, the more unfortunate surprise would probably be if it didn't keep quite as many as expected. For example, someone who chooses to keep snapshots of the last 1 day, the last 1 week, and the last 1 month, and then finds out when they need to recover something that they only have the single most recent backup of less than a day ago, might feel misled. I think it would be safer to err on the side of retention. I would favor replacing "last N" with "preceding N" or "previous N" to be clear about it and then making sure that the behavior matches that. - Derek
![](https://secure.gravatar.com/avatar/cdfdc7382d7a988a6574e78e3bb6310c.jpg?s=120&d=mm&r=g)
Dear Christian, I didn't answer outright, because I gave it a hard and long thinking about this session. Please #define N 3. In short, In week 48, I expect - W45, W46, W47, if BiT didn't take this week's backup. - W46, W47, W48, if BiT took this week's backups. However, it's a bit more complicated than that. I have questions: - What will be the effect if I take daily backups (which I do). Then I expect this week's individual backups (W48), and W47, W46 & W45 as weekly backups. - What will be the effect if I take weekly backups only? Then I expect W48, W47, W46. Since the numbers are tunable, the algorithm looks very complicated to approach and get it correctly. What I propose to implement is a slightly more complicated manner: - There's should be a well defined model of how backups are grouped. - Maybe BiT should consider having "buckets" of backups (daily, weekly, monthly, yearly). - A backup should migrate to the broader bucket (daily -> weekly -> monthly -> yearly) buckets as soon as it's kicked off its current one due to age. - The first entry of a bucket should allow snapshots merged into it until its rotation time comes. - If you keep 4 daily backups, the 5th one should move to weekly, and should accept new backups merged into it for a week, then rotated. - If you keep 6 weekly backups, 7th should be demoted to monthly, and should accept more weekly backups until it's a month old. The model is not complete, and somewhat complicated, but I think it can be iterated to something which makes sense and easier to implement and maintain. I'll try to polish it and report under the GitHub bug, too. Cheers, Hakan On 11/26/24 6:23 PM, c.buhtz@posteo.jp wrote:
Hi Hakan,
thank you for your thoughts.
Am 26.11.2024 12:26 schrieb Hakan Bayındır:
"I'll have N weekly copies at all times"
OK, but is the current week a week no matter if it is complete or not? ;)
Let's imagine your setting says 3 weeks to keep. Today we do have week number 48 in this year. What do you think? Will BIT keep the 3 backups for week 45, 46 and 47 or 46, 47 and 48?
I would like to know what you expect as a user. I now what BIT is doing or try (because of bugs) to does in this case.
Christian _______________________________________________ Bit-dev mailing list -- bit-dev@python.org To unsubscribe send an email to bit-dev-leave@python.org https://mail.python.org/mailman3/lists/bit-dev.python.org/ Member address: hakan@bayindir.org
![](https://secure.gravatar.com/avatar/62653e4429d3606c32fa59311cfdeae1.jpg?s=120&d=mm&r=g)
I think that there are two problems here: 1) the language is confusing, and 2) the behavior of the code is buggy. As for the bug, I once started to investigate it deeply, but I got stuck and never got to the bottom of the problem. My current state of knowledge is documented here: https://github.com/bit-team/backintime/issues/1094 Cheers Michael On 26.11.2024 11:56, c.buhtz@posteo.jp wrote:
Hello, Please take this as a user survey and answer out of your gut ignoring technical details or wording and translation issues.
Looking onto the "Auto-remove" tab you do see this option.
"Keep one snapshot per week for the last N weeks"
Does this take the current week into account even if it is not yet complete (e.g. today it is Thursday)?
Regards, Christian _______________________________________________ Bit-dev mailing list -- bit-dev@python.org To unsubscribe send an email to bit-dev-leave@python.org https://mail.python.org/mailman3/lists/bit-dev.python.org/ Member address: foss@michael-bueker.de
![](https://secure.gravatar.com/avatar/148a51ecc80531586e1f5530d64b988a.jpg?s=120&d=mm&r=g)
FYI I opened a meta issue about "Revise Auto- and Smart-Remove feature" currently not sure where it will go. <https://github.com/bit-team/backintime/issues/1945>
participants (5)
-
c.buhtz@posteo.jp
-
Derek
-
Hakan Bayındır
-
Michael Büker
-
zahlman