I been noticing a trend of people nuking tags (for example, forum #217092), then saying that the alias would be unnecessary because it was nuked.

I believe that the alias would still be valid for completeness (to avoid the wrong tag from appearing in the future). Any thoughts?

Every accepted alias and implication adds strain to the servers. If a tag is only tagged a few times and the chance of it being tagged again is low enough, fixing it yourself is more practical than waiting for an alias to be accepted.

BlueDingo said:
Every accepted alias and implication adds strain to the servers. If a tag is only tagged a few times and the chance of it being tagged again is low enough, fixing it yourself is more practical than waiting for an alias to be accepted.

The only time I'd see the servers flexing their muscles in the slightest is when someone tries to use the tag that was aliased. It's not like alias/implication checks are a regular server task that must be ran every 5 seconds... Maybe I'm wrong, but from what I saw when d621 was a e6 clone, I didn't notice any server lag from the aliases or implications.

Granted, each individual alias'/implication's effect on the servers is negligible and only kicks in during searches and tag edits, but their effect accumulates when there's a lot of them.

From what I can see, there are ~9720 accepted aliases and ~7100 accepted implications, many of which are triggered constantly and some of which are unnecessary (tagme_if_anything_important_is_missing -> tagme? Really?). These numbers could easily be doubled if every possible alias/implication was added to the lists when you take into account missing character -> copyright requests, things with many names (see double_bass), common spelling mistakes and lots more.

Basically, you can't make an alias/implication for everything. Sometimes, it's easier to just fix it yourself.

BlueDingo said:
Every accepted alias and implication adds strain to the servers.

Citation needed on this. I don't know what e621's code looks like, but searching through aliases sounds like it should be a constant-time operation.

There might be merit to other arguments for avoiding aliases. But I won't believe the performance argument until I see some hard data (e.g. benchmarks) to support it.

There's no way I could obtain that data, but it makes sense if you think about it.

More aliases accepted = more aliases checked during each tag edit and search = more processing power used.

The actual amount of processing power needed is unknown to me. You'll need one of the developers to tell you that.

Like I've said, it's more of a staff issue. Approving aliases takes time, and the admins always have better things to do than to approve something that could've been easily handled manually.

Which is why there's such a large backlog of those.

Genjar said:
Like I've said, it's more of a staff issue. Approving aliases takes time, and the admins always have better things to do than to approve something that could've been easily handled manually.

Which is why there's such a large backlog of those.

Okay, that makes sense. Not so much a processing-demand issue, but rather a human-processing issue.

Last I knew, the alias/implication system was also a little messed up in that it didn't always carry through to all relevant posts correctly (though IIRC an admin said that was a timeout issue, so probably mostly occurs for particularly popular tags)

Can confirm, it's not a processing time issue since aliases are cached anyways.

In addition to the (admin) time it takes to process them and potential bad aliases down the road, it makes it much more difficult to rename a tag if we ever decide to; aliases need to be moved manually, which is a huge pain when there's a bunch of aliases on it that never get tagged anyways.