Why does HN timeout so fast?

https://news.ycombinator.com/item?id=2733376

by kaushalc • 15 years ago

22 14 15 years ago

news.ycombinator.com

If I dont use HN for a few mins it expires and clicking on next link shows an error message saying link expired. Why is this? Its a pain that I need to go and reload the main page and the click next to come back to the page i was on.

cpayne624• 15 years ago

Yeah. It's pretty whack

MattJ100• 15 years ago

Glad you mentioned it, I was beginning to think it was just me.

pokoleo• 15 years ago

I believe that there was a post about this before.

IIRC, the way that HN is set up, it re-distributes the "fnid=NSgN9i46gR" parameters at the end of the string to the first/second/third/etc. pages.

Reddit (a similar example) does this, but it also does it in such a way that the parameters "?count=50&after=t3_ihoin" fit into a SQL(/alternative DB) select statement.

This would probably be a better way for HN to do it, but they probably have a reason for why it's currently done this way .

wkearney99• 15 years ago

Could you at least change the error page to include a link back to the main 'new' page?

Rust• 15 years ago

It does seem to vary a lot between about 2 minutes and 5 minutes of inactivity. Like I can read or even nicely skim 30 articles in 2 minutes...

Coincoin• 15 years ago

My guess is that it depends on the amount of traffic.

Edit: I remember a few years ago it would take 30 minutes to an hour before it expired. Maybe it deletes the oldest tokens first.

pasbesoin3• 15 years ago

A while back, when the site was rather bogging down, pg made some changes to memory management (as I vaguely recall from one of his comments posted at the time). I believe part of this is more aggressive discarding of memory, and that part of the result was to succeed, or fail, more quickly without progressively tying up the server.

So, overall the server works better. When there's load, part of this include bailing sooner on your troublesome request. ;-)

In your case, this may include no longer having memory/context for stale pagination links. I deal with this by using a browser extension to load several pages of links at once. (Without a browser extension, one can achieve the same results by opening those links manually in new tabs and then working one's way through those tabs at one's leisure. I'd advise not overdoing this, though; no need to burden the server with requests you will seldom/never use.)

(I may be wrong as to the memory management. Again, this is just my vague understanding/recollection.)

zoudini• 15 years ago

What browser extension do you use?

pasbesoin3• 15 years ago

https://addons.mozilla.org/en-US/firefox/addon/google-docs-p...

At the time I adopted it, IIRC I read through the code enough to determine it was not malicious. I've been perhaps lax in not checking all of the updates.

I see its max version is listed as 3.6 . I'm actually still on 3.6; when I go to 5, I guess I'll see whether it still works with max version overridden.

pasbesoin3• 15 years ago

Sorry to reply twice, but my previous reply was made quickly and I incorrectly recalled the wrong context. (I also recently mentioned using an extension to view PDF's in Google's Document Viewer.)

The extension I was referring to in my grandparent comment is:

https://addons.mozilla.org/en-US/firefox/addon/repagination/

(Note that there is another, older "Re-Pagination" extension that was abandoned. IIRC, this one picked it up and updated it.)

PLEASE BE CAREFUL if you use this extension. If you do not choose a limited page count, it is easy to unduly load or overload a server. This include the HN server, and such request behavior will get you cut off or banned (rightly so; please work to preserve HN resources).

I use it, infrequently, a time or two a day to load the first few top pages of results, which I then work through at my leisure.

I also use a bookmark to transform all the link href values to open in new pages/tabs. That way, I don't lose the browser-generated multiple-page page (as it were) by forgetting to right click or shift-click to open a link that appears on it in a new tab.

jones1618• 15 years ago

It is brain-damaged that both Reddit and HackerNews do this.

Their next page URL's should either be page based ("show me page 4, even if page 4 has changed") or based on the next article ("show me the page that contains article 1234"). You might see a few duplicates are miss a few articles but you wouldn't get "unknown or expired link" errors.

shii• 15 years ago

http://news.ycombinator.com/item?id=17675

http://news.ycombinator.com/item?id=2677600

donnaware• 15 years ago

It seems ironic that the same people that sit around and obsess about what the latest groovy language is that you supposed to be using to make your web site with can not seem to make a very good web site for themselves. But then, I am addicted to Hacker News, so what does that say about me?

theoa• 15 years ago

And me too...

Why does HN timeout so fast?

Related Stories