I’m glad you all like NewsBlur enough to continue working on convincing me, but it’s not going to happen anytime soon. The extra work has to do with the intelligence training and the fact that I would have to fetch twice as many stories from the database, do twice as much counting, and it would end up doubling the amount of work I have to do in feed processing, effectively slowing down the entire UI for everybody, not just you. I’m happy to offer refunds because I only want highly satisfied and happy customers. There are numerous alternatives, although none that I like as much as NewsBlur, that are happy to give you unlimited unreads for the tradeoff of having abysmal speed.
Does the server store the full content for all articles? just unread?
Where does the huge impact on your servers come from? What kind of impact are you seeing now? Perhaps by understanding that we can find an answer that works better for everyone.
For me losing the read / unread state is effectively losing the article. I have over 400 sites and 4300 articles. I do intend to read/skim each and every one. Once it is marked read an article is lost to the pile, and if it is marked by the system and not me, I may not even know to look for it.
I have busy periods and slow periods and often save catching up on my feeds for slow periods.
I am a digital pack rat. I don’t want to let any article go until I’ve had a chance to read it.
Samuel, thank you for continuing to talk with us about this. We wouldn’t be so adamant about our concerns if Newsblur weren’t so amazing.
Of course the server keeps the stories. We’re talking about unread status, not stories themselves. You can still read the stories, you just don’t have an unread indication. And if you’re out of town, then you already know the feeds you want to read and can just continue scrolling down until you hit the stories you saw before you left.
But if you have to read every story, then you are already reading them faster than every 2 weeks. And if you’re not, then you are well outside the bounds of what I can reasonably accommodate. You are a power user, and as a power user I gave you a great way out: install your own hosted version of NewsBlur.
I personally don’t care about intelligence training at all, would it be easy to disable on a per-account basis? And I might take you up on that refund offer 
I wrote this above as a reply, but it bears repeating:
Remember, NewsBlur keeps stories both read and unread. We’re talking about unread status only, not stories themselves. You can still read the stories, you just don’t have an unread indication. And if you’re out of town, then you already know the feeds you want to read and can just continue scrolling down until you hit the stories you saw before you left.
If you have to read every story, then you are already reading them faster than every 2 weeks. And if you’re not, then you are well outside the bounds of what I can reasonably accommodate. You are a power user, and as a power user I gave you a great way out: install your own hosted version of NewsBlur.
2 Likes
but Samuel that’s exactly where, i think, you are wrong about the usage of people requesting more than 2 weeks : we are not power users !
We just want to be sure that if we don’t think about reading the news during 3 weeks, we will still see our 1-post-a-week feed counting as unread … I don’t care about getting all the Apple Insider news of last month but i would like to be sure i don’t miss my small sites’ news …
In 2013 that’s quite strange to have to keep an Evernote note with the last date of reading RSS feeds just to be sure not to miss one … that should be called unread feature 
The unread indication is the whole reason I’m using a feed reader, otherwise I could just as well keep a folder of bookmarks.
Think of feeds for TV and online shows, webcomics, or any other custom feed that only updates rarely, but when it does you definitely want to know about it. This doesn’t just apply when you’re out of town, sometimes you just don’t have the time to read a really long essay, or watch an hour-long video, but still want to keep it around for later, without having to save each story individually.
Again, the real problem is that you’re even using the date to decide which stories to delete. Deleting 2 week old data for performance reasons is silly when the size of that data is a handful of unread states (with these kinds of feeds). What’s wrong with switching to a limited *number* of stories instead? You could even use a feed’s post rate to normalize the limit, until you’d end up with the same server load as today, but a system that would keep more older stories for small feeds, and slightly less for high-traffic feeds.
That does seem like an excellent compromise. I have some low-frequency updating feeds that require dedicated time set aside to read. Under the current approach I lose these stories.
Alternatively, as suggested above, if disabling intelligence training was an option to maintain an extended unread count I’d be all over it.
It’s not the only thing that people are complaining about; try commenting on a shared article that was shared without an initial comment
It doesn’t matters that is only the read count, the whole point is that this behavior is not expected and at least it should be configurable.
Read count tracking is one of the main features for a RSS client. If I can’t trust it wether I read something or not it loses a lot of its value.
What exactally is that impact? Is it bandwidth, storage, CPU time? This may be obvious for you, but I have never run NewsBlur on my own hardware and I have no clue.
As I can understand, in the past you made the architectural choice of using a centralized global constant to set the deadline we’d have to delight the unread flag feature, and now it’s too hard to change.
It seems you alone work on NewsBlur development. That’s why in another post I talked about lack of a development community.
I totally understand you not wanting to spend resources on free accounts.
But I can’t understand why you go to the point of offering refund to unsatisfied customers. Is this unread issue so expensive that you’d rather lose income than spend some more money in resoures? Is the resources consumed so expensive that subscription wouldn’t pay it and you’d stay in loss?
Of course none of us want you to bankrupt and shut down your servers. We don’t want NewsBlur to follow FeedDemon’s path. We want you to profit from it, so that it can be your main income and you dedicate all your work time and your talent on it.
Try to understand me. 1 year ago you was thinking about a lifetime subscription business model. We explained how that was a terrible idea and you dropped it. Couldn’t it be the same situation here?
Another point. I used to have a feed with 50-100 posts every day. It’s be 1000 unread articles if I just layed it there. I wasn’t able to follow, so I just deleted it. This kind of feed is already torturing your servers even with 2 weeks deadline. That’s 1 situation.
Another situation is websites like mine, that stay months with no post. I have a few feeds in this situation, that rarely have something new. But when it does, I wanna read it. This kind of feed could freely had its articles unread for months and would barely touch any resource.
Based on those situations, my suggestion is that you’d put 2 deadlines: time and amount. Let’s say free accounts will be allowed to keep articles for 1 month and 200 articles. If any of those deadlines are reached you just totally delete oldest ones.
For premium accounts you increase it for 6 months and 1000 articles. If somebody wanna have more freedom, just pay for the resources. Then you can create segment subscriptions, if somebody needs even more time and amount they pay more for it.
2 Likes
Here’s a solution (albeit a weird one)–just have all your analytics run on only two weeks of posts. So if it is unread, but it’s outside the window, then it won’t slow down the UI. And, if keeping the read / unread state for all those posts is too much (I would have a hard time believing that, but it could be…), then just keep a per feed “most recently read post”–so I could always jump to the most recent post that was read (note, not the oldest unread post–so if you skip around, you lose all the posts before the most recently read post, but if you read in order it works).
Seems to settle the server and UI load and gets everyone the abilities they look for in a reader.
1 Like
Is there a technical reason for it?
I’d host my own version, but as far as I know there is no option in the android app to use another api url
I missed earlier where Samuel said “The extra work has to do with the intelligence training and the fact that I would have to fetch twice as many stories from the database, do twice as much counting, and it would end up doubling the amount of work I have to do in feed processing, effectively slowing down the entire UI for everybody, not just you.”
That explains why unread articles create such a surprising load on the server. Would it be possible to make the training just look at the last 2 weeks? Training for many is a secondary feature to keeping unread counts, even though it is one of the distinguishing features for Newsblur.
ahhh now I got it! The problem is the amount of processing needed to handle Intelligence!
Well, then the issue is directly related to amount of articles. The query will limit it by a LIMIT N or a WHERE post_date > now + 14 days. Now I understand why the centralized global constant, any query can simply take it and print in the SQL, instead of having to handle it in a per account, folder, feed, etc basis.
Kinda troubling. But I still think that premium accounts should have some love here. If you’re making the query, WHERE statement already have info about the feed/folder and the account, so it’s not that expensive to also have a field for days related to it. It could just be account-wide.
Maybe, instead of having a value for each account, have a bool flag for it being premium or free, and 2 global constants, and a simple if will solve it. I suppose you can give premium accounts at least 60 days. And think about it, lots of ppl want the resource, so it would give NewsBlur an advantage over other web-based RSS Readers.
The 2 things I’m waiting to get my premium account is the solution for this issue and upping saved stories.
Another thing to look into is how Newsblur currently stores user-specific bits of metadata for posts much older than two weeks, and syncs those bits across contexts, evidently without stress to the servers. I have in mind saved/unsaved and shared/unshared.
The storage is not the problem, it’s that the setting for the number of days to keep is also used in various other places, such as the intelligence trainer.
“Sold” more than one year ago – I think this needs to keep getting bumped, please.