3 ms·

> a lot of feeds don’t have pubDates I don't see how this would significantly impact the cost. The main cost is that you need to remember all entry IDs to know

> a lot of feeds don’t have pubDates I don't see how this would significantly impact the cost. The main cost is that you need to remember all entry IDs to know which you have seen before. However you should probably do this anyway because published dates aren't very reliable. Plus if you are storing user comments then you probably want to keep the entries around for a long time anyways. > polling or archive retrieval are expensive Polling isn't very expensive. It is just a HTTP get. Many feeds also support conditional requests so you just get a 304. If you have enough feeds that becomes expensive you can play with the update time, poll feeds of active users more often, slow down polling if a feed doesn't post often... I also don't know how archive retrieval is expensive. You only really need to do this when a new feed is added (and even then you only need to do it if you want history). Maybe if the full "first page" of the feed is new you can look back a bit into the archive but this is so rare that it probably isn't worth doing. I guess for pagnated (not archived) feeds you should technically check all pages each time but so few readers even support pagnation at all that I would consider a feed that doesn't put new entries on the first page broken and wouldn't bother support it. TL;DR I don't think this service would be very expensive to run. If you get enough users and feeds to start being concerned about the cost of polling you will have enough resources to deal with it. Source: I run a small RSS reader service.

pipeline_peak 4y ago
How entry id’s serve as an alternative to pubDates? Also by entry id, do you mean RSS item? Most of my concern comes from seeing NewsBlur, Feedly, and The Old Reader resort to premium services. Throttling users unless they pay, now that’s not a game I want to be apart of.
- kevincox 4y ago
  My only guess for how pubDate would affect the cost was using it to identify previously seen entries/items. If you can rely on pubDates being stable and monotonically increasing for new entries you can just remember the last one instead of a number of IDs. But that is likely an incredibly small cost. How else do you expect to make money other than premium services? Even if cheap hosting is not free. And likely the cost of staff is by far the biggest expense of these companies. Premium literally means something that costs. The only other major business model I am aware of than charging your chstomers is to sell their data or attention. Personally I prefer the freemium or paid model. If you are going for freemium than faster update rate sounds like an incredibly fair compromise if that is increasing the cost of serving a user. Much better than many artificial limitations.
  - pipeline_peak 4y ago
    > How else do you expect to make money other than premium services? I don’t, I want something that can host for users as cheaply as possible. I’d make it open source and be fully transparent about the fat client design and whatever else. I pretty much want this site, but every user has an RSS feed, can see each other’s, and a front page with rankings.
    - kevincox 4y ago
      Even this site is funded, it is effectively an add for Y-Combinator and it also hosts job listings for YC companies. I guess "make money" is poor wording but privately funded often fails in the long-term especially if a site becomes popular.

pipeline_peak 4y ago

How entry id’s serve as an alternative to pubDates? Also by entry id, do you mean RSS item? Most of my concern comes from seeing NewsBlur, Feedly, and The Old Reader resort to premium services. Throttling users unless they pay, now that’s not a game I want to be apart of.

kevincox 4y ago
My only guess for how pubDate would affect the cost was using it to identify previously seen entries/items. If you can rely on pubDates being stable and monotonically increasing for new entries you can just remember the last one instead of a number of IDs. But that is likely an incredibly small cost. How else do you expect to make money other than premium services? Even if cheap hosting is not free. And likely the cost of staff is by far the biggest expense of these companies. Premium literally means something that costs. The only other major business model I am aware of than charging your chstomers is to sell their data or attention. Personally I prefer the freemium or paid model. If you are going for freemium than faster update rate sounds like an incredibly fair compromise if that is increasing the cost of serving a user. Much better than many artificial limitations.
- pipeline_peak 4y ago
  > How else do you expect to make money other than premium services? I don’t, I want something that can host for users as cheaply as possible. I’d make it open source and be fully transparent about the fat client design and whatever else. I pretty much want this site, but every user has an RSS feed, can see each other’s, and a front page with rankings.
  - kevincox 4y ago
    Even this site is funded, it is effectively an add for Y-Combinator and it also hosts job listings for YC companies. I guess "make money" is poor wording but privately funded often fails in the long-term especially if a site becomes popular.