My Initial Thoughts on Bluesky's AT Protocol | Hacker News Reader

My Initial Thoughts on Bluesky's AT Protocol

havenweb.org

30 pointsmawise34 comments

Open articleView on HN

I was confused. I thought they were talking about the protocol you use to call a BBS from your modem: “AT DT 555-1234”

Tell me you're old (at least Xer age) without telling me you're old.

I'm not that old and know of the AT command protocol, it is used a lot in modems for embedded systems

AFAIK it's still used cell modems to communicate between the baseband and the application processor.

OK, Generation X BBSers are now officially rechristened "Generation AT".

I'll take that

It's used in a lot of small embedded devices. I'm in my twenties (for reference) and I've used them a lot of ESP8266, SIM900A, SIM7000 programming.

It's non-ironically a very good system that "just works" most of the time without having to deep dive the documentation.

I wish they would have named it something else.

https://www.developershome.com/sms/atCommandsIntro.asp

I feel a bit bad about that. I'm 36 and I'm the oldest on the team, but Usenet was before me (which seems to be why it's widely known? though I see it's more than that). Googling brought up the Hayes commands but we had no idea it was this strong of an association.

It's probably still in great use with telemetry devices.

I worked at a small company that developed GPS + Cellular devices. Most of the firmware I worked on processed AT commands, either off the RS232 serial bus, or the network adapter. These devices were either deployed to remote locations for monitoring, or on vehicles for tracking purposes. Basically all the commands to get data or execute commands were customized AT commands.

It is still used in ESP8266 based ESP-01 WiFi boards that people connect to Arduino and the like to make it an IOT device. AT commands that it support allows you to connect to wifi and then establish TCP/UDP.

As far as I'm aware most modems use these commands, even very new and high end ones.

I actually used AT commands to call BBSes and... ignore these people. We shouldn't turn civilization into a museum. It's fine to reuse old names for new things.

Same here & concur. Even the AT commands weren't typically called "AT protocol".

It's even more valuable for people to be challenged, & upgrade their skill, at context-based disambiguation than to further ossify short names. The preeminence of the `@` symbol in social-network addressing – from email through Twitter & everywhere – also makes this a well-chosen name.

+++ATH

For better or worse, AT commands are still alive and well, along with mainframes, COBOL, and OS/2.

They'll probably outlive me. They'll probably outlive all of us.

AT stands for ATTENTION. In command mode, it configures the modem for data transfer. It is an excellent name for a social network protocol.

> On the whole there is a lot of good stuff, but I worry about the ability for it to stay an open standard. Today Google has dominant browser market share with Chrome, so they get to dictate new web standards just by building them into Chrome.

I think this concern is mostly obviated by the use of content schemas ("lexicons"): https://atproto.com/lexicons/atproto-com

My concern is how many of these schemas will need to be implemented for true compatibility. At this stage at least, it's still better than implementing ActivityPub.

I briefly read their doc, I think being able to migrate my data from one host to another is really great, but I'm unclear about the concept of repository.

according to the doc, it seems that data stays in repositories. A repository is a user-signed storage, representing all data created by a user.

does that mean, this protocol is only suitable for twitter/blog type of applications?

I'm thinking about forums/BBS/Reddit, on which data (a thread) is not owned by a specific user, but by a forum board?

Thanks for the writeup. It’s an interesting protocol, but I’m not sure account mobility and portability is quite as you paint it in your post. The main problem is that, like email and mastodon, most users will sign up through a service, rather than self-host. I wrote about this in another comment thread here:

https://news.ycombinator.com/item?id=33260045

You're 100% on this. That seems to be the whole point of DIDs, but under today's internet I think BYOD and domain ownership is a better solution. That said, none of the registrars seem to be working towards end-user domain ownership.

TakingNames.io is a cool project that works a little bit in this direction, they provide an OAuth-type flow to grant an application permission to update your DNS records on your behalf.

Domain-ownership is why we're including did:web as an option. That's the fully self-sufficient model. The did:plc method is a way to give people a solution if they don't own the domain.

As someone who lost access to their domain for 2 days due to massive registrar incompetence (don't bargain bin shop for your cute tlds, folks)... No. Please god no.

If everyone uses a common protocol then they are limited to what the protocol accommodates and there is really no way to distinguish yourself from everyone else using the same protocol.

Wouldn't this apply to email with POP and SMTP? And yet, users have clearly distinguished between good (Gmail, Hey) and bad (AOL, Yahoo) clients.

Yes, that's the idea.

My initial thoughts on your thoughts went over my head

> Imagine having an email address that you could move from Gmail to Yahoo Mail to Outlook, and all the emails would keep reaching you.

You mean like...email? Or Jabber? Or Matrix? Or SIP?

I've moved my email a half dozen times, moved my Jabber twice, and SIP three times while keeping the same address. Sometimes self-hosted, sometimes commercial. Matrix is newer to my life so I haven't had a need to move it yet, but whenever that time comes I see no reason I won't be able to do so.

This isn't rocket surgery, it doesn't require new protocols, it just requires people to be willing to spend a small amount of time and money to own their own addresses.

I'm not really familiar with Mastodon or the rest of the "Fediverse" but as far as I understand those they should be equally portable.

Presumably you understand how DNS records (and more generally the entire domain name system system) work. Do you think your system would work for someone who did not have any of this background knowledge? I feel like this is still a significant barrier and at least part of the reason why most people use gmail/hotmail/yahoo etc.

> Presumably you understand how DNS records (and more generally the entire domain name system system) work. Do you think your system would work for someone who did not have any of this background knowledge? I feel like this is still a significant barrier and at least part of the reason why most people use gmail/hotmail/yahoo etc.

I do, but I didn't need to. For my email for example I was one of those early adopters of Google's former free tier for consumers with their own domain. My DNS provider, like most, had a wizard where I just told it my email was hosted with Google and they set up all the correct records for me, MX, SPF, etc. The same sorts of wizards would be trivial to set up for other major services supporting self-hosting if they were popular.

Those who need more custom setups could still edit the records themselves, but for most people using a major hosting provider it doesn't have to be hard.

There's also of course an obvious opportunity for some sort of meta-SRV record that'd allow a user to tell their DNS host that they want to set up service X with provider Y.tld and the DNS host could look up the relevant records and set it up accordingly.

What I'm saying is it's not an insurmountable problem, or even really that big of a technical challenge, to make control of your own online identity easy. It's just not in the commercial interest of enough of the big players.

I bet they used Domain Connect[1]! That's the mainstream cross-service domain management protocol, but they require each service explicitly register with each domain registrar. I also don't think any of the companies involved are super invested in it right now which is a shame.

[1]: https://www.domainconnect.org/

Server providers could help their customers with it.

We are where we are, same place as around 20 years ago. Custom personal email domains are a very niche item. Service providers have no incentive to bother serving this niche, and in fact are disincentivized from making moving easy in general due to their business models.

There are many services that will host your email with your own domain, and they hand hold you on what and how to change DNS to make that happen.

It's not free, but then, why would it be? If you want to own your id in a decentralised system, you should expect to either host it yourself or pay a fee. If you want free, then you need to understand that you pay for that service by not owning your id (and probably get bombarded with ads etc...). I don't think that's unfair, and at least there is space for the former, contrary to the current social "networks".

Maybe I'm misunderstanding you, but you don't need to understand DNS to set up email with a custom domain. Providers like Fastmail are happy to handle all of that for you.

Reply on news.ycombinator.com