[gxqoz]

Yeah I have 32k saves and hit the same problems with search being extremely unreliable. About 5 years ago quotes stopped working in search. Trying to find "The Grapes of Wrath" would return all instances of "of" and "the." You could sort of hack it by searching for the most distinct word (maybe "Grapes") if you already knew exactly what you were searching for. I long suspected there was some architectural change they made on the backend that broke this and they didn't want to admit in support articles. Perhaps the Mozilla legal department determined that having a text copy of all articles in their database was some legal risk and they moved to just having the URL and maybe the title (this would also explain why "permanent copies" disappeared).

Anyway, as the 32k articles indicate, I was a power user of Pocket so part of me is sad it's going away. But they've really been checked out since maybe 2019 with regards to any real support for this product.

[chii]

> having a text copy of all articles in their database was some legal risk

the risk should've been the same with google's index, and yet they're dandy!

I think it's more easily explained by incompetence. Esp. when stop words like 'of' and 'the' are somehow included in the index. These are almost trivial to remove prior to indexing (any decent indexing library, such as lucene, would have a prepared list of stop words filter, and it's not like you even need to do any work to have it!).

[kamarg]

> the risk should've been the same with google's index, and yet they're dandy!

Sure it should be but reality says Google has many more and probably better lawyers so the risk is clearly different.

[sshine]

> suspected there was some architectural change they made on the backend that broke this

I think Mozilla specialises in this. At some point, having many tabs open on the Firefox iOS app would eat all the memory and lag out the phone. This problem even came and went a couple of times over the years.

It’s an unloved child being held captive for the money it earns.

[GuestFAUniverse]

IMO search is garbage in all Mozilla products.

E.g. Thunderbird ignores potential matches in quoted mail text. That's utterly useless if one remembers a certain mentioning by the other side. Plus, now and then repairing the index suddenly leads to matches -- when is the right time to repair? I don't now -> always if it's seriously important...

[E39M5S62]

Thunderbird search is bad enough that I just open up the Gmail website to find an email. At this point I don't even know why I use a local email client, except maybe 25 years of muscle memory.

[bsder]

See, now "local repository search" is one of those things that you would think that somebody would use one of the LLMs for.

Alas, that would be "useful" but not Unicorn Lottery Ticket Useful and so will never get implemented.

[alias_neo]

Is something like Apache Solr (a search index) well suited for something like this?

I've deployed and used it at work for searching specific, well-specified bits of information, but I don't know how well it would work on large chunks of text like articles etc; I assume this is its real purpose and it should fit, but I'm guessing.

[retinaros]

Just pgsql is enough. Even a chache db or sqlite do full text search