jay@thecapacity's dashboard

@Robin & @Arne

Both of you make good points, although I'm still in the camp w/ Robin (i.e. I'd like the data too). :)

I think Arne's point about being able to track the ratio of ping's to pongs (sorry that's how I envision them) is actually useful and you're right that it doesn't have to be exclusively one or the other.

I can only imagine the capacity issues you have to account for, but I do feel inclined to point out that with with the "webhooks" approach there's very little of the parsing that has to occur. I'd think it should just be a hash lookup, though of course the task of queuing and relaying is still there.

Anyway, I think I must apologize for trying to reply to each in return and creating the spaghetti approach to a discussion ;) – jay@thecapacity, on April 21, 2009 18:39

It sounds reasonable, I just wanted to try to better illustrate my "thought process" and why I think there would be more info if you catch the initial "ping" rather then the return "pong" :D

I really want the data to come across in addition to the ping so I can shave off a 2nd step of gathering it. But I can understand the size constraints... i.e. your comments to Robin below.

As a thought of XML vs. JSON, maybe that could be a configurable parm when I set up the endpoint? i.e an option that says "deliver this to me as <>" – thecapacity, on April 21, 2009 18:34

I just wanted comment on this to clarify how I think I'd end up using this service;

1) I'd setup notify.me to alert me when there's a message, say on an RSS or email queue or whatever your support.

2) When I got pinged via notify.me I'd say "thanks for the alert" and then I'd simply go check what queue / service I'd registered to send to my URL

Now that I better understand what you're trying to do I think that probably sounds "aggressive" or "anti-competitive", i.e. I'm exploiting your service and not "giving back".

I really don't mean it to sound like that (I'm a big fan of supporting people to make possible the services they provide!!) but I thought I'd try to explain my "interpretation" in case it helped better present my perspective. – thecapacity, on April 21, 2009 15:16

@Robin, thanks that's a much clearer explanation then anything I could have come up with! – thecapacity, on April 21, 2009 15:11

@Jonathan thanks for the link I've actually been using smtp2web for a while not but it's really not a strong solution for production (that's not a criticism just an observation).

@Arne - Absolutely right that the transport munging is one critical aspect. I wouldn't know your business as well as you, but as I mentioned in an earlier post it seems splitting hairs to say "I'll notify you a message exists in this queue / transport type" but not just go ahead and deliver some digest (full or partial) of that message, i.e. the message itself can serve as notification.

I wouldn't think that you track the URL shorted version would be the best way to get a usage breakdown, i.e. my program may not follow that URL most / all of the time. I'd expect the only "certain metric" you could rely on is the act of delivering to my endpoint.

Webhooks is certainly "plumbing" but I wouldn't overlook it anymore then I'd say RSS, Email or even twitter & facebook are plumbing. I'm less sure I understand the nuances behind your statement that it's less "a direct consumer delivery" but I think that's what I see webhooks and RSS (in theory though it's of course reversed in practice) as.

Anyway, don't mean to come off as a contrarian but it's definitely a need I have, i.e. to convert from polling to something more responsive. I know there's Gnop (sp?) but can't seem to find it and having checked out Switchhub too I can tell you it's just a service vacant of any providers. – thecapacity, on April 21, 2009 15:10

My view on what webhooks say is that there is no "question" about what it's referring to because it's be definition open... i.e. it doesn't specify what's being transported or to where, other then it's via an HTTP verb.

I think your "wiring" gives me a canonical URL to your servers which I will have to lookup and dereference to get to the actual data. Not necessarily bad if you could provide meta info on that, e.g. historical information, i.e. change control on a REST resource, that is assuming I wanted it for my context.

Sounds like for now you're just going to; (a) alert my URL handler then (b) proxy the URL request I make to the actual resource

I'm not sure I see the value in (b) but that might just be my desire to optimize a step (and a point of failure) out of the equation. – thecapacity, on April 20, 2009 16:09

I think I agree with Robin at this point. If you can't commit to delivering me a string representation of the message, then don't deliver me much at all (just the service URL to go check). – thecapacity, on April 20, 2009 16:04

My concern with key/value pairs is that, for example, the python email library supports concerting a string into an email object. In my experience that's how most libraries are used to inputs, e.g. BeautifulSoup and simplejson parse strings.

If you hand me key/value pairs I'm going to have to shoehorn that input to make it acceptable for just about every library out there. – thecapacity, on April 20, 2009 16:04

Robin, I think the point you make about "notification" vs "message delivery" is an interesting one. I tend to think that if you're going to notify me that there's a new message(s) waiting you might as well just deliver the message itself in most cases but if it's just a ping to wake me up to go poll RSS that would work too. – thecapacity, on April 20, 2009 16:02

I do think spamming request parms is ugly but I think the fact that you "will not be able to act as a delivery agent for arbitrary message bodies" is a concern.

I think starting with "we support the generic anything case" seems a much better approach because it scales the fastest. Me having to rely on you guys supporting email, rss, atom, webhooks, etc... all on your own time (and priorities) is a huge issue. As an example, I have a twitterbot ready to go but the fact that I can't get the SMTP messages translated (reliably) into POST data has stalled me for weeks. – thecapacity, on April 20, 2009 16:01

@Arne

I'm not sure why you'd want to explicitly "POST an XML body"... What I think you'd want it to post whatever the service gave you, i.e. I don't think you should be a translation later or add any type of meta info except perhaps in the URL itself.

Not sure if that made sense but as a principle I'd say "leave the message alone" and then you're not forced to try and reinvent any wheels at all, nor do you end up with something nasty like an XML encapsulated JSON message.

My 2nd rule of thumb is JSON for everything over XML, since we are talking JSON here but I realize that's like saying VI rulz ;)

Any meta information you really wanted to pass on should be in the request parms, e.g. &from=twitter@twitter.com

Hmm, I must have missed the email support. I saw it as an output but is it also functional as an input? I guess my biggest need is is the URL post anyway.

Maybe you can get together with switchub.com. They don't support email inputs but they have URL outputs. That's the marriage! ;)

Glad it's on the radar. Not to sound too contrarian but I'd think email and RSS as inputs would be more useful then an API since the former would mean I can instantly connect to my existing services rather then trying to get them to support your API or writing my own translation scripts.

Anyway, I'm definitely excited by a web with pipes!

I want this to send the notification to a URL I specity... ala "webhooks" or smtp2web

It doesn't do it on the regular site...