I didn't realize that this was a use-case anyone was interested in, but it sort of makes sense. I've added a parameter you can pass to RH.org to deal with the gravatar pull on my side, so you don't have to.
This adds a bit of serverload, since I need to make a bunch of requests, but it's not THAT bad.
If you pass gravatar=yes, it will make a pull to the gravatar URL for that address. If something exists there that isn't the default, it will issue a 301 over to the site.
Otherwise, it will return the robohash you requested.
It also passes the size param over to Gravatar, just to be nice ;)
Interesting point. Wavatars are Shamus Young's original work.[1] Identicons[2] and monsterids[3] are Scott Sherill-Mix's work. I don't think those guys were shooting for "ugly", but I agree an ugly or nondescript avatar could incentivize users to upload a custom one. And that may have played into Gravatar's decision to support these avatar types.
Minor correction: Identicons were created by Don Park who open sourced a Java implementation. Scott adapted it for WordPress and gives credit to Don credit for the idea in your link[2]:
"A couple weeks ago I made a WordPress plugin to generate unique monsters for each commenter. Don Park came up with the original idea for representing users with geometric shapes. Since I already had the framework in place I thought I’d make a WordPress plugin for the original geometric Identicons."
Ah. But I'd expect think that they'd limit it to only images, or do some sorta safety checking. I recognize the performance concerns, but it's too easy to add a malware redirect onto their trusted url.
Very nice, I was planning to implement a similar thing with for 'anonymous' ip-based human readable usernames to identify users without forcing them to create profiles.
An excellent bonus - my robohash looks exactly like Bender from Futurama :^)
I'm curious how many variations robohash uses. It seems either you've got a very wide variety of custom graphics elements, or only a few elements and have to draw a very wide number of images to fit in them.
Each two digits of hex is 255, and I count 16 pairs, but can think of only 8 elements in the photos (eyes, ears, nose, mouth, head, body, arms, background).
If it's those 8 elements and they're derived from four digits in the hash, then you need 65,535 or so different ears.
I'm guessing the solution to this is to reduce the hash further... so you get more repetition where two texts produce the same robot, but apparently they've not made it too bad.
How far did robohash reduce it? Or how far should one reduce it? How many images does a robohash art set have?
I didn't realize that this was a use-case anyone was interested in, but it sort of makes sense. I've added a parameter you can pass to RH.org to deal with the gravatar pull on my side, so you don't have to.
This adds a bit of serverload, since I need to make a bunch of requests, but it's not THAT bad.
http://robohash.org/colin@sq7.org?gravatar=yes
If you pass gravatar=yes, it will make a pull to the gravatar URL for that address. If something exists there that isn't the default, it will issue a 301 over to the site.
Otherwise, it will return the robohash you requested. It also passes the size param over to Gravatar, just to be nice ;)