Something that's come up in discussions a few times recently is how important it is to have good default settings. Even users who are quite technical and involved don't always explore which settings are available, and that's totally fine—they shouldn't need to. The default setup should be as good as possible, with changing settings mostly for specialized cases.

One particular place on Tildes where this isn't currently being done well is for the "mark new comments" feature, which has always been disabled by default. I think it's one of the best features on the site and makes it much easier to follow ongoing discussions here than on other sites with similar comment systems, but overall, not many users have enabled it.

For example, Tildes got some attention on Hacker News again yesterday, and about 80 new users have registered so far from that. Only 9 of them enabled "mark new comments", even though the welcome message strongly encourages it. Looking at longer periods of time, this seems typical: only about 10% of users ever enable it.

As it says on the settings page for the feature, my reason for disabling it by default was out of privacy concerns. However, I've been doing some review of the data that Tildes stores lately and realized that this was kind of misleading and inaccurate. Because I have HTTP request server logs and some other related data (which is all only kept for 30 days), I effectively have topic visit records from the last 30 days for all users anyway, whether they have the feature enabled or not. The data is more convenient to access for users with the feature enabled, but it's available either way.

Because of that, and because the data will be very useful to combine with some of the upcoming changes I mentioned in the last ~tildes.official post, I'm planning to enable this feature for everyone. Here are the general plans:

• Data about which topics' comments pages a user visits will be stored (for 30 days), along with when and how many comments were there at the time. This enables displaying which topics have new comments since your last visit, and marking those new comments.
• There will no longer be a setting to disable this, but you can still choose whether previously-seen comments are collapsed when you return - the same as the existing checkbox on that page for "Collapse old comments when I return to a topic".
• I will probably implement some sort of "stop informing me of new comments in this topic" feature (separate from the new Ignore one) to stop having the info about new comments in a topic showing up for you.

Please let me know if you have any thoughts or concerns about this. If nothing major comes up, I intend to make this change later this week.

This is a privacy-related update that I've always intended to implement on Tildes, and I finally spent some time on it this week.

Keeping eternal records of everything that every user ever voted on is some of the most sensitive data that sites with a voting system have. Your voting history says a huge amount about you, your interests and opinions, and can even serve as a decent proxy for showing what times you were active on the site, what posts you were reading, and how long you spent reading the comments on each of them. In exchange for these major privacy implications, you get the tiny benefit of being able to tell which old posts you voted on (if you even go back to old posts).

So now, to match up with Tildes's general approach of deleting as much sensitive data as possible after 30 days, the voting on posts closes when they're 30 days old. After a post's voting is closed, the records of which individual users voted on that post are deleted, but the count of how many votes there were is kept. So old posts will continue showing their same "scores" exactly the same as before, but there will be no record of which individual users cast those votes.

However, this isn't a purely positive update: the main downside is that the voting does need to be closed (otherwise there would be no way to prevent people from voting again after their first vote is deleted), which prevents the occasionally useful ability to vote on old topics or comments. Overall though, voting on older posts is extremely rare, with less than 1% of the votes on Tildes ever made on something that was over 30 days old at the time of voting.

When the "delete old sensitive data" job runs for the first time after this update later today, 97% of the voting data in the database will be deleted. That's a massive decrease in the amount of sensitive data the site is retaining, and something that most sites would never consider doing, because of the value of that data for behavior analysis and ad-targeting.