Stride - XMPP and API support and timeline.

[Steven Skoczen]

Hey Atlassian team,

I'm Steven, the author and maintainer of Will , one of the bigger Hipchat bots (It's listed on your integrations page.)

I'm in the middle of a big upgrade, and in the process, am seeing things here that really worry me.  Namely, that the Hipchat XMPP protocol seems to be changing, has invalid SSL certs, and that the timeline and migration path to Stride isn't clear.

And more concerningly - all of that comes as a surprise.  As far as I can find, it's not documented anywhere.

I'm at a point where I might need to decide to drop hipchat/stride support completely in the upcoming release, because I'm not sure what spec to build against.

Can you please provide:

1. A description of current and planned changes to the chat.hipchat.com XMPP protocol?

2. A description and timeline of any moves from that protocol to Stride's?

3. An answer on whether Stride will support XMPP, or another secure, non-webhook-based protocol (this is important for people running bots on their own systems without public-facing endpoints.)

4. An overall narrative of how you see current users moving from Hipchat to Stride.

From the outside, it looks like Atlassian is spooked by Slack, and is winging it in trying to ship a replacement as fast as possible.

That makes sense - but one thing you might be missing is that a lot of your current customers are still on Hipchat specifically because of the integrations they've built.

We have a few core developers on the project whose companies specifically still use Hipchat only because Will runs their entire devops stacks.  To keep those customers, you'll have to provide a path for those teams to move to Stride without disruption.

I wish you all the best of luck with Stride, and I'd love to support it (and continue to support Hipchat) in Will.

I look forward to your answers, and being able to do so!

-Steven

[Tanguy Crusson · Accepted solution]

Hi @Steven Skoczen,

Tanguy from the Stride API team. I want to start with an apology - I just realized there wasn't a clear answer to give you a path forward in supporting Will for Stride and this is not acceptable. I'm hoping this answer will clarify the situation.

So here goes:

• Your observations are correct: Stride does not support XMPP today. I want to be really clear that we don't have plans to build this support. Integrations using XMPP will have to be rebuilt using the Stride API (based on REST/Webhooks). 
• Building support for bots like Will should definitely be possible using the Stride API. In fact, we recently shipped a Hubot adapter doing just this. You can find it on npm, and on Github.
• We'd love to work with you, and developers of other popular bots like Lita or Errbot, to help you get on Stride. We currently work with hundreds of bot developers on the Stride API with the goal of supporting the best of the bot world in Stride.  

If you'd like to chat in more details, please contact us on strideapi@atlassian.com and we'll set up a call.

Cheers,

Tanguy

[Tim Howe]

Mar 21, 2018

I want to start with an apology

Hey, great start.

I just realized [emphasis mine] there wasn't a clear answer to give you a path forward in supporting Will for Stride and this is not acceptable.

I'm hoping this is a regrettable attempt at humor since the credulity you're expecting from the community is neither fair nor polite.

I want to be really clear that we don't have plans to build this support.

I want to be really clear that this gives the lie to the explicit promises you made that all HipChat features would be fully supported.

These promises -- obviously vacuous at the time, granted -- came from the top and were unambiguous and unconditional.  You don't get to drag paying customers onto an unwanted and untried service on the strength of trust in your care for their needs when in the same breath you blithely dismiss which features those customers are allowed to find important.  Not to mention the sheer principle of the thing: when Atlassian promises something you knew from the start you had no intention of providing, what does that say?

If one is forced to migrate away from HipChat, why should they choose anything from Atlassian?  Indeed, why should they?

P.S.

If you'd like to chat in more details, please contact us on strideapi@atlassian.com and we'll set up a call.

The eagerness you have to remove this discussion from public eyes is disappointing but not surprising.

[Tim Howe]

HAHAHAHAHAHAHAHAHA

[Steven Skoczen]

Update on Oct 26th 2017 - I've heard from the Atlassian team via the Ask-the-CEO link Boris suggested above.

My current takeaway is that things are still quite rapidly being built, and are subject to significant change (which is good news.)

At the moment, I've heard that:

1. XMPP support will be completely dropped in Stride, and that

2. All connecting services will need publicly addressable endpoints, and need to use webhooks.

I've also been pointed to the WIP documentation, which I'll link to in a comment to in case of overactive spam filters here.

I'm continuing to talk to the team, and emphasizing that dropping XMPP and not replacing it with something similar looks (to me) to be a disastrous decision, especially since all their competitors have a similar feature, and frameworks like Will rely on it.

The best I think we can do is continue to have open, positive, and honest conversations as customers and developers on the direction we'd like the product to go, and hope we're heard. At the end of the day, it's Atlassians' call.

[Steven Skoczen]

WIP documentation: https://developer.atlassian.com/cloud/stride/learning/messages/

[Tim Howe]

That's quite unfortunate.

Sounds like they're forcing customers over to a system that's not even all the way built yet?

Not to mention the contradicting information over continued feature support and when they will be forcing users to migrate.

Time to start looking at competitors, as you said.  However if things are still in flux I'll hope with you that they choose to replace the missing capability.

[Steven Skoczen]

What they've said to me is that there isn't a forced migration or specific timeline:

"HipChat users will be moving to Stride group-by-group, on their own timetable. We've just become the process of inviting folks to migrate. Because the Stride API and apps are different (and more capable!) that HipChat integrations, customers will need to install and configure new applications in Stride after they migrate."

That's not a specific timeline, but I'm not sure a timeline exists yet.  I'd read that as "HipChat will be supported indefinitely", but it's certainly not a bulletproof guarantee.

[Tim Howe]

"more capable" LOL

[Tim Howe]

@Steven SkoczenI'd like to believe that but I believe they've already locked admins out from making any changes, and there's this FAQ:

What if I want to move back to HipChat?

Starting soon, we will begin to upgrade all HipChat Cloud teams to Stride. You are welcome to continue to use your HipChat Cloud instance until you upgrade. After upgrading, HipChat Cloud will remain in 'read-only' mode so that you can reference your API configuration in HipChat Cloud and receive integration notifications for those apps that are not yet in Stride.

[lauren]

Hey Tim! We haven't locked admins out from making any changes, you will have to option to schedule your own migration to Stride. But yes, eventually all Hipchat Cloud customers will have to move to Stride. Like @Steven Skoczen said, there isn't a specific timeline.

[Dave]

We don't care whether you use XMPP internally, or IRC, or carrier pigeons. That's your business.

What we *do* care about is having to migrate to stripe, and then suffering the slow and painful education process as you figure out how to re-implement the XMPP functionality we need (presence, history, multi-user chat, security and authentication), but without using the XMP Protocol, and while your developers learn the hard way that distributed systems are, well, hard.

It's clearly going to be less work and less pain for us to migrate to an alternative setup that does use open standards, than it will be to migrate to stripe, and enter your walled garden of the future.

 I'm sure stripe will be full of cool things, but from a business perspective what we need more than anything else is group text chat and secure fully indexed history of a few channels. Everything else (video & audio chat, for example), we already have options in use today.

If your API is literally unable to support basic group chat, then you already have a problem. And if it *does* support it, then adding basic XMPP is not going to be a major hassle for you internally. XMPP means almost any existing service or developer can bang out an integration or a bot that scales wickedly, using reliable well tested code on any platform.

Atlassian: your job here is to convince us somehow that the community having to re-implement all of the XMPP functionality repeatedly in javascript/nodejs, in python, in elixir, in php, in ruby, C++ etc is somehow more efficient for the whole hipchat community, than Atlassian simply providing one robust bridge between your new APIs and the existing world standard for chat.

Up to you to decide if moving to a walled garden is long term a better strategy than the open and integration friendly setup you already have with hipchat.

[lauren]

Hi Dave,

Thank you for the feedback. Clearly you are very passionate about XMPP and this space, and we are glad that you reached out to us. 

Know that while we can't do anything to change this right now, we hear you. We'll keep an eye on this opportunity as we progress.

[Dave McKenzie]

Nobody's passionate about XMPP, we're just passionate about not reinventing the wheel and wasting a lot of engineering time.

[Nic Brough -Adaptavist-]

I don't think I'd call it an "opportunity", more of a "necessity". 

Hipchat did a (not great, but) useful implementation of XMPP that enabled a LOT of integration and sharing. 

If Stride is to succeed, it needs to provide something similar - either something new that is so great, people will move off XMPP to use it, or XMPP so, as Dave says, no-one has to re-invent things.

[Dave]

@laurenthanks for the reply. To be clear, I'm not passionate about XMPP, I'm just bringing 30 years of experience to the table. XMPP works, on low power devices, and on desktops. I don't need to wait for Atlassian to ship a native client for my obscure platform, I can just plug in one of 3 or 4 open source tools.

I've seen this repeatedly with Novell servers, Microsoft Windows, innumerable desktop Office suites, the list goes on. Companies go bust, product priorities change, new versions with new but incompatible features arrive.

Open standards always win out in the long term. There's nothing incompatible between having a core that does better and fancy things with bots, and remaining with an XMPP/IRC bridge. Slack's versions of these aren't great, and you have an opportunity here to milk that.

 Again, go ahead and use non-XMPP in the backend by all means, but just make a reasonably compatible bridge sometime in the next year or so.

I should mention that I followed Boris' advice below, and contacted the stride PM via CEO office and got a curteous reply saying that XMPP compat is not on the roadmap. I suggest a few more of you do so, ultimately the customer's voice is heard.

[Nic Brough -Adaptavist-]

Slack is dropping XMPP (and IRC) soon.

[Ryan Gillan]

I was just about to point out this exact same thing (re Slack).  Atlassian aren't the only ones.

A few confusing (and amusing) comments here.  XMPP on low-powered devices?!  Speaking specifically about mobile, it's one of the worst protocols to use regarding battery use, and countless adaptations have been since done to attempt to make it more "connectionless".

I see mentions of walled gardens too -- don't understand where that's coming from, considering the (new) APIs for HipChat and (I think Stride) are fully open and documented.  This does not equal a walled garden.

[Tim Howe]

Thanks for asking this @Steven Skoczen. This isn't the first time it's been asked but Atlassian conspicuously hasn't answered so far, and unfortunately in the past whenever they've tried to avoid a straightforward answer to the questions of paying customers it's turned out that those customers are about to be left in the rain, and strangely the pattern has been the least notice for the most impactful changes.

Very well-written points and it's an excellent example of trying to understand Atlassian's position while pointing out the real-world impacts. Ourselves, we built a lot of workflows around Atlassian services but have had to steadily move away as one after another has compatibility or key features broken. HipChat is yet another example.

Unfortunately it's becoming less individual incidents and more of a pattern with Atlassian, one that seems to be accelerating. This is especially concerning as they largely specialize in configuration management.  A couple of years ago, I could speak positively of the Atlassian experience; I don't know that I can do that now.

Atlassian has encouraged organizations to build key workflows around HipChat, and compatibility is a key attribute of any platform.  They have an example now to set a precedent and show how much they value dependability when people are counting on them.

[Steven Skoczen]

For sure, Tim.  I'm having some conversations via email with the Atlasssian team as well, and I'll keep this ticket up to date with what I learn (I have a strong transparency bias.)

[Marc Linster]

Adoption will be much more difficult if Stride does not support integration with Pidgin, Adium etc. Asking folks to install and watch yet another client app is challenging - especially in an open source environment where XMPP, IRC etc are well established

[Boris Berenberg - Atlas Authority]

I’m not sure if you will get the answers you need here. You may try contacting the Stride PM and seeing if they can help answer some questions. FWIW, as a partner I too don’t know the answers to your questions.

[Steven Skoczen]

Thanks, Boris. 

Any recommendations on how I get in touch with the Stride PM?  Don't see an obvious way via the community or website.

[Boris Berenberg - Atlas Authority]

Official way would be to use the contact the ceo form and request to have your questions answered by the Stride PM. They may turn you down, but it’s the least backchannel route you can take.

[Steven Skoczen]

Excellent, will do that now - appreciate the point in the right direction.

[lauren]

Hi All,

So sorry for the delay here, I wanted to make sure I had all the right information from the Product team. 

Unlike HipChat, Stride is not built using XMPP and will not support an XMPP interface.

We believe that Stride is so much more than just a chat app, and XMPP alone could not support the all of the features we envision for Stride. 

That means all users need to use one of the official Stride clients, and generic XMPP clients will not be supported. 

That also means that any integrations dependent on XMPP will need to be rebuilt. Integrations built for HipChat are hugely valuable. In building Stride, however, we realized that we needed to make some fundamental changes in order to provide the kind of capabilities and dev experience that we wanted. We understand that it will mean work for developers to update integrations to support Stride, but we believe the long term enhancements will be worth it.

Thank you for your patience and understanding - this wasn't a choice we made hastily. 

If you'd like early access to our new API, please sign up here: https://developer.atlassian.com/cloud/stride/

[digitalfiz]

Having XMPP support has very little to do with supporting all the things you want to be able to do and more to do with supporting a protocol that opens your platform to more options for integration. XMPP is fully capable of sending pretty much any kind of structured data to/fom clients and servers so while you may not feel the urge to build all your functionality into XMPP it would still work for allowing bots to integrate and provide the same functionality they did on HipChat. I don't think bots really care about voice and video chat or whatever super rich experiences you guys are adding to Stride...

[Jonathan Mell]

what @digitalfiz and @Dave have said are correct. Build a reverse proxy for the bots to make everybody happy. 

[Tim Howe]

HAHAHAHA @laurenhow do you feel about telling customers to spend resources on porting to Stride APIs?