How much does external API integration suck? (survey) | Hacker News Reader

How much does external API integration suck? (survey)

trove.polldaddy.com

22 pointsAntiFreeze11 comments

I answered your questions, but I think you should probably change your questions. You guys have very specific assumptions of what a "project" is (when does a project begin and end? What if I consider an API integration a project in itself?) without making that explicit, so the answers you get are going to diverge wildly because of that ambiguity.

I agree completely. We were aiming for really rough data to validate some rough assumptions of the space, and didn't expect the massive response to the survey.

Given a second chance, we'd be much more specific about these questions. Thanks for taking the time to answer!

   On these integrations and maintenance now versus last
   year (i.e. how much more time are you spending this
   year on these integrations as compared to last year):

Sorry, but this is poorly worded, and doesn't allow you to say you're spending less time integrating APIs this year than last.

(The answer selection is a percentage scale.)

Who tracks this? I mean other than just a gut feeling.

And how much of it is integrating an API vs. doing it yourself. (eg. You might be spending twice as much time integrating APIs and half as much time doing it yourself)

Also, does external mean a webservice or a module you didn't write yourself. (eg. Facebook vs. Win32)

We were aiming for gut feelings, which is why we went with percentages in the poll.

These are definitely important questions that the poll didn't take into account, that we'll most likely need to address in the near future.

As for external services, we were really looking for anything you did to bring in external data, which could be anything from writing the connectors yourself to just loading a module and passing through some API keys.

The biggest problem IMHO with the APIs is inconsistency, for instance on the FB graph API on some calls it gives you a 302 redirect instead of returning JSON.

so instead of { image: "http://foo.com/image.jpg } you have to catch the Location header and write it into your object.

Next up would be how to represent a date in JSON which seems to vary between an time_t cast to an "int" and a string in ISO-8601 format.

Lets not even get started on XML and attributes vs. elements.

You're absolutely right. We totally screwed that one up, and it just didn't make sense to change it mid-poll.

Honestly, in all the work we've been doing (and the work our friends have been doing), it appeared obvious to us that integration was consistently getting worse and we were curious how much worse.

It is now abundantly clear that we were completely off base here. Guess we'll have to run another survey soon with much better answer choices given what we've already learned in less than a day.

Thanks for the feedback and taking the survey!

Hey all, just wanted to give you a heads up. There are three of us working on this, I'm on the west coast, sick as a dog, and highly medicated, and the other two are on the east coast and already asleep. We'll reply to you all, but realistically it'll be in the morning!

Thanks for helping out!

I tend to build projects around an API - not just integrate an API service into something for authentication or the like.

after spending all day in a scrum planning meeting, i really wanted to fill out the survey but as soon as i read "how much of your time is spent" i ran away screaming.

seriously, though, i'm interested to see what comes out of this. what apis are you targeting? social apis? payment processing? provider apis, like libcloud?

Hah, thanks ;-)

We're trying to reduce the friction between applications and user content. In other words, to do away with all the unnecessary plumbing that most people build every time they want to access a user's 'stuff.'

That plumbing looks like a ton of CRUD-like operations that repeat themselves (auth, query, normalize into your own data dictionary, health check, etc.).

We're at first targeting content APIs (the 'stuff' mentioned above), trying to create a query mechanism so you can do broad queries with a user, without needing to know or care about what endpoints the user's stuff resides in. i.e. get all photos tagged dog, regardless of whether those photos live in facebook, flickr, picasa, or some other web service you've never heard of.

We're having a blast and even have the photo bit work.

Reply on news.ycombinator.com