The discussion revolves around the idea of having 750 categories in a Discourse community. pfaffman is importing a Ning community about oil rights and is considering using categories to organize content by county and state. Despite reading about the potential issues with tags and categories, pfaffman believes this is the best solution.
codinghorror suggests that while having 750 categories may not be a disaster, it may be overwhelming for users. codinghorror proposes using tag views as a homepage and muting all tags by default, allowing users to un-mute the ones they care about.
Other users, such as angus, suggest directing users to their preferred county category and restricting the loading of categories in the site model to improve performance. Mittineague and ron_jeremy propose alternative solutions, such as using parent categories or creating sticky topics in state categories.
sbernhard shares a similar issue with having many main categories and subcategories, and asks about potential performance problems for common users and admins. pfaffman shares their experience with a similar site, suggesting that performance is okay, but the 99th percentile can be problematic.
Overall, the discussion highlights the potential challenges of having a large number of categories and explores alternative solutions to improve user experience and performance.
I’m importing a Ning community about oil rights. The belief is that those who own rights in one county don’t care a whit about what’s happening in other counties.
I’ve read It’s Time We Talked About Tags and paid attention to, I think, every topic about the issue over the past ten months.
There are about 10 general interest categories, and then about 20 state categories and each state has 5-50 counties for a whopping total of something like 750. I’ve tried to figure a way around it, but have failed to find a better solution.
One thing I’ve considered is using groups to limit access to categories so that people would see only the categories they care about then they hit /latest or /top (which I can still do).
The problem with that scenario is that it’s important that search engines be able to see all of the categories.
It seems OK on the test site (4GB DO droplet). /categories loads in 1012.8ms and /latest loads in 726.
It’s not a disaster if you can keep it under 1000, but I am not sure users will be able to comprehend that many categories. I dunno. Try it, see how it works, and report back in 6 months?
It’s good that you checked the page load times, though, that’s certainly advisable.
So why not dump them into their tag view and have that be their homepage? I’m not sure why that would be so problematic? Do they really need security here?
Your willingness to edit people’s comments never ceases to amaze me, but that’s not important now.
That’s something of a relief.
I’d think the same, but they’ve been comprehending them on Ning, which I find hard to . . . comprehend.
Well, partly because I’m not that familiar with tag view, perhaps I should do some re-reading.
No. They don’t need security; using groups to limit access was my attempt at giving people just the trickle they want from the fire hose. They need to be able to see only the counties that they care about. If you’ve got minerals in Oklahoma, you don’t care about the goings on anywhere else.
And when the people in Whosit County Kentucky generate 200 posts a month about their dirt, there is a bunch of noise for most people.
Is it possible to mute all tags by default and then un-mute the ones you care about? That might solve the problem.
Just don’t ever visit the homepage; come in at your preferred county. If this is the case, the only time you’d ever need to see the homepage is on your first visit, yes?
Heck a static HTML page with a map that people clicked, linking to the appropriate tag for that county, would be the best homepage in this case, yes?
So maybe the feature request is to make a tag route your preferred homepage, or even a category cc @sam we already allow people to select latest, categories, or top as their preferred homepage.. why not a category or tag (or group!), at that point? why not?
I have a topic that is a list of very county category, so that’s a start, but you might care about 10 county categories plus the 10 top-level general interest categories.
Yeah, this is how I plan to deal with this issue on a similar setup. Direct people straight to their ‘home’ category.
If I hit performance problems with 1k categories, I’ll also restrict the loading of categories in the site model somehow. It loads all categories by default whenever you load the site.
@pfaffman I’m very curious to hear how this turns out actually.
It looks like you have a category per county. With state code as a “comma code” at the end.
Is it too late to rethink hierarchy? For me, having parent categories would help. If not by state, some other geopolitical boundary. I might miss out on something “hot” happening in a far county. But would I be all that interested in keeping up with three time zones away?
A hornet’s nest is the only thing that comes to mind when i think of managing 750 categories. Hell, Craigslist likely covers the entire country with less than 750 categories.
Could you create one category per state and create sticky topics in each category to cover the counties? You’d also have the freedom to use the category for ‘general state talj’ etc.
If that’s true, it’s because they skip over a whole bunch of places. 750 is just counties that have mineral rights. And about 30 states don’t have any.
Again, they managed 750 “groups” on Ning, so that’s what they’re used to (which doesn’t really make it any less crazy, but that’s another matter).
월간 페이지뷰가 약 50만 회에 달하는 사이트를 운영 중입니다. 카테고리가 수백 개나 되는데(대략 300개 정도인 것 같지만 수정: 814개입니다). 시스템 사양은 2코어 {4스레드} CPU와 16GB RAM을 갖춘 비교적 빠른 환경입니다. 성능은 괜찮은 편입니다. DigitalOcean의 8GB CPU 최적화 인스턴스를 사용했던 것으로 기억합니다.
시도해 보는 것 말고는 다른 방법이 없다고 생각합니다. 각 그룹 간의 상호작용이 적다면 멀티사이트(multisite) 인스턴스로 전환하여 여러 데이터베이스로 분산할 수 있습니다. 단일 사이트를 SSO 서버로 공유하는 방식도 가능합니다.
중앙값 기준 성능은 꽤 좋은 편이지만, 99퍼센타일(99th percentile) 성능은 상당히 나쁜 편입니다:
