Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

URLs are UI

hanselman.com

Post reply on HN
Print view View original

1–10 of 96 posts

Re: URLs are UI

#2

Post by userbinator »

Yes, I absolutely agree that good URLs are extremely useful, although personally I don't really like the long-string-of-text pattern, especially if it isn't actually significant. A short identifier is useful when sharing them offline: "HN article 14723409" or "YouTube video AQcSFsQyct8" or "forum thread 5705591"

For a blog, news, or other chronological content, I'd like to see a timestamp of some form. If it's rigidly hierarchical content, then a hierarchy (of which I should be able to remove 'subdirectories' to see the parent content) makes sense. Otherwise flat IDs are OK too.

Too bad browser developers seem to love hiding or mutilating them...

Re: URLs are UI

#3

Post by seagreen »

Absolutely true, and also a problem.

The fact of the matter is unskilled people are always going to be designing websites. This is OK!

However, it does mean that we should require them to do as little as possible so they have the best chance of getting it right. Asking them to design a second UI right from the start on top of the first HTML/JS one -- and then telling them never to change it(!) -- is a little much.

Instead of URLs, websites should have UUIDs to identify each resource. They should also have metadata describing what that resource is. The metadata (eg "articles/urls-are-uis") should be able to change without breaking links to the resource. Browsers should be intelligent enough so that when you hover over a link you see the metadata, not the UUID.

(This has one downside which if if you link to thing X and the UUID is later changed to point to thing Y, it may look like you linked to something you didn't mean to. This can be trivially fixed by including a "what the metadata was when I made the link" field in links along with the UUID)

EDIT: I'm actually not set on UUIDs specifically, they're a little long. Any random, non-meaningful identifier is fine.

Really I'm just saying that points 2 & 3 of the article are so good we should have made them the default (and perhaps only) option from the start.

Re: URLs are UI

#4

Post by chrismcb »

userbinator wrote: post #2

Yes, I absolutely agree that good URLs are extremely useful, although personally I don't really like the long-string-of-text pattern, especially if it isn't actually significant. A short identifier is useful when sharing them offline: "HN article 14723409" or "YouTube video AQcSFsQyct8" or "forum thread 5705591" For a blog, news, or other chronological content, I'd like to see a timestamp of some form. If it's rigidl…

A string of text is much easier to remember then a bunch of digits. Exorcism if there are more than 4 digits in a row

Re: URLs are UI

#6

Post by saurik »

chrismcb wrote: post #4
userbinator wrote: post #2

Yes, I absolutely agree that good URLs are extremely useful, although personally I don't really like the long-string-of-text pattern, especially if it isn't actually significant. A short identifier is useful when sharing them offline: "HN article 14723409" or "YouTube video AQcSFsQyct8" or "forum thread 5705591" For a blog, news, or other chronological content, I'd like to see a timestamp of some form. If it's rigidl…

A string of text is much easier to remember then a bunch of digits. Exorcism if there are more than 4 digits in a row

The string of text, if you are seriously trying to remember it (which I seriously doubt... do you really try to type in those sentence-long title slugs from memory?!) is subject to domain-specific forms of corruption, in the same way direct quotes from people or TV characters tend to be: you subtly change the grammar or replace a noun with a synonym with which you are more familiar.

Regardless: I am quite serious... do you seriously try to remember and type, from memory, URLs with title slugs?

Re: URLs are UI

#7

Post by morecoffee »

Which of the following URLS are clearer?

    1a. https://foo.com/
    1b. https://www.foo.com/

    2a. https://foo.com/search/products/couches/color/red
    2b. https://foo.com/search?products=couches&color=red

    3a. /posts?recent
    3b. /posts?recent=
    3c. /posts?recent=true

    4a. /post/1234/products-couches-furniture-red-cheap-recent-ikea-tags
    4b. /post/1234/
I can think of good reasons for picking each of these, and even good reasons for picking their alternatives. URL design is tricky to get right because realistically you can't change it later.

Re: URLs are UI

#8

Post by csours »

See also the pit of despair known as SharePoint.

example.com/sites/MySite/MyList.aspx

can be functionally the same as

example.com/sites/MySite/MyList.aspx?GUIDGUIDGUIDGUIDGUIDGUIDGUIDGUIDGUID=GUIDGUIDGUIDGUIDGUIDGUIDGUIDGUIDGUID&GUIDGUIDGUIDGUIDGUIDGUIDGUIDGUIDGUID=GUIDGUIDGUIDGUIDGUIDGUIDGUIDGUIDGUID&GUIDGUIDGUIDGUIDGUIDGUIDGUIDGUIDGUID=GUIDGUIDGUIDGUIDGUIDGUIDGUIDGUIDGUID&GUIDGUIDGUIDGUIDGUIDGUIDGUIDGUIDGUID=GUIDGUIDGUIDGUIDGUIDGUIDGUIDGUIDGUID

Why, SharePoint, why?

Or even worse, Angular SPAs

example.com/page#nice_page_you_have_there/shame_if_you_tried_to_share_it

As a tester, I always check:

1. What happens if I go to "example.com", "http://example.com", "https://example.com"

2. Sharing URLs - at least you should get a landing page, but I've seen 404s from shared URLs

Post reply on HN