I’ve been working on this for a while, but let’s make it official: I started a little Tumblr-like microblog about software craft and quality!

[doekman@mastodon.nl]

@mwichary True to that. Reminds me I always wondered why no API documentation existed on Dropbox/Google Drive/Nextcloud URL’s. I had to reverse engineer them. Apparently API's should be REST…

[sylvhem@eldritch.cafe]

@mwichary It makes me happy to see that others care about URL.

[mwichary@mastodon.online]

@Ember Yeah, I agree that part is a bit muddy where every scheme has pros and cons. This one seems to indicate there is more hierarchy than there actually is…

[mwichary@mastodon.online]

@mez Good question. I don’t know! I naturally put it at the front but that might be because I only saw it that way. Interesting point about autocomplete!

[csilverman@mastodon.social]

@mwichary The idea of URLs as user interface elements is such a good take. I've seen some people use URLs as design/communications elements as well, like Jessica Hische:

https://www.jessicahische.is/thinkingthoughts

https://www.jessicahische.is/working

https://jessicahische.is/anoversharer

I love that approach. Modern browsers and preview cards often obscure URLs, but people still see these things; printed materials, links in emails, etc.

[nelson@tech.lgbt]

@mwichary this clean URI scheme is also why FlickrClient.py could support 68 methods of the Flickr API in just 48 lines of code. The core magic:

def __getattr__(self, method):
  def method(_self=self, _method=method, **params):
    _method = _method.replace("_", ".")
    url = HOST + PATH + "?method=%s&%s&api_key=%s" %
          (_method, urlencode(params), self.api_key)
    try: rsp = xmltramp.load(url)
    except: return None
    return _self._parseResponse(rsp)

[polarbirke@indieweb.social]

@mwichary Much agreement!

There is plenty of prior art, some examples that I particularly remember:
- 2010: https://warpspire.com/posts/url-design/
- 2017: https://www.hanselman.com/blog/urls-are-ui
- 2018: https://alexpounds.com/blog/2018/12/29/four-cool-urls

[mwichary@mastodon.online]

@polarbirke Thanks for sharing!

[cogdog@cosocial.ca]

@mwichary I love the Flickr url pattern. You could not find a collections link with knowing the pattern.

I am fine with the unique id as shorter url. Also knowing where it is in a downloaded photo allows the secret url trick to find the original page https://cogdogblog.com/2015/10/flickr-trickr/

[mwichary@mastodon.online]

@cogdog Whoa! This would’ve saved me so much trouble throughout the years.

[jimniels@mastodon.social]

@mwichary fwiw I wrote some of my favs a while back

https://blog.jim-nielsen.com/2023/examples-of-great-urls/

[mwichary@mastodon.online]

@jimniels Thanks! Added here, maybe it helps someone! https://unsung.aresluna.org/mailbag-urls-as-ui/

[splorp@mastodon.social]

@mwichary @cogdog One pattern I use a lot to locate my own images is for tags.

https://flickr.com/photos/splorp/tags/blue

[mx@typo.social]

@splorp @mwichary @cogdog Slightly offtopic, but I still use `/photo.gne?id=` URL to find where I downloaded certain photos years ago: https://www.flickr.com/photo.gne?rb=1&id=3399133066

[luke@typo.social]

@splorp @mwichary @cogdog look at that patina on that bus design. Just don’t look at the poorly applied window tint.

[splorp@mastodon.social]

@mx @mwichary @cogdog Just the fact that those legacy Game Neverending URIs still function after all this time is mind-blowing.

[splorp@mastodon.social]

@Luke @mwichary @cogdog Those air pockets make my teeth hurt.

[mwichary@mastodon.online]

@splorp @mx @cogdog Cool URIs Never End

[dokas@mas.to]

@mwichary @splorp @mx @cogdog Just wanted to pipe up and mention that while the photo.gne trick isn't advertised it's known and so useful I even have a Launchbar search plugin for it.

[imsop@tech.lgbt]

@mez @mwichary

Off the top of my head...

ID first means the URL can survive truncation of the slug part. Particularly good if you can keep the prefix short.

Slug first means URLs sort in alphabetical order by their slug. Maybe more suitable for small numbers of items, like a blog?

ID first is also marginally easier to construct a URL by hand - you generally don't need to add a fake slug, just end the URL at the ID and it still meets the pattern

Random example of slug-first that happened to be in my history:
- canonical URL: https://www.hollywoodreporter.com/tv/tv-news/stargate-revived-prime-video-new-series-1236431118/
- doesn't work with only ID: https://www.hollywoodreporter.com/tv/tv-news/1236431118/
- works if you know the format and add a fake slug: https://www.hollywoodreporter.com/tv/tv-news/yeehaw-1236431118/