"Web scale" is useful only to corporate hypercapitalism.

Federation on decentralized open protocols is its antithesis.

@tindall this doesn't make sense outside of the social network space. think about some large-scale public service, or even just an ordinary news site; you kind of need those things to scale and federating wouldn't particularly help

@aeonofdiscord @tindall why shouldn't a news site federate? NNTP did federate. Also, a connectivity problem today between my region and eg. USA shouldn't make it impossible for me to view yesterday's news from USA.

@Wolf480pl NNTP facilitates discussion groups; the word "news" in the name seems to be a historical accident. retrieving yesterday's news when the server is down is a caching problem; if that's federation then any site that uses a CDN is "federated" and the term doesn't mean anything

@aeonofdiscord nah, what I mean is, each news site should cache news from other news sites. And any news site should be able to join. CDNs aren't like that, because you have to pay them for them to cache your stuff.

@Wolf480pl even under that system, each individual site would need to be "web-scale"

@aeonofdiscord no, it wouldn't.

@Wolf480pl how are you proposing to transfer the bits

@aeonofdiscord Let's assume N people want to read your news, once aday, and they're scattered around M different regions/communities/etc. In a centralized scenario, they'd need go directly to your site, so your site would need to handle N requests a day. In a federated scenario, they all view the news through the news site that's closest (for some definition of distance) to them. Each of sites handles N/M reqs from users, and M reqs from other sites a day. N/M + M < N

@aeonofdiscord if M ~= sqrt(N), then you have 2*sqrt(N) requests a day instead of N

@aeonofdiscord Let's assume there's 4 * 10^9 ~= 4 * 2^30 = 2^32 people with internet access in the world. sqrt(2^32) = 2^16 ~= 64 000. That's a totally different scale than 4 billion.

@Wolf480pl readers have preferred news sources, and outlets have preferences over what kinds of news they carry. you won't see perfectly equal distribution of load across all nodes; most likely you'll see a power-law distribution

@Wolf480pl also obviously this is ignoring issues like figuring out how news producers get paid for news

@aeonofdiscord no no you're doing it wrong.
We haven't yet figured out how a for-profit could ever make something federated. So either you assume it's a non-profit, or you expect it to become centralized, sooner or later.

@Wolf480pl I'm talking about payment for doing work, profit is a whole different set of questions

@aeonofdiscord well, then you need a federated tip jar I guess?

@Wolf480pl I'm not convinced that setup would be sufficient to replace the news industry as it currently exists

@aeonofdiscord
>industry
Of course not. Mass-produced news won't work in that scenario.

@aeonofdiscord I'm currently writing a federated dictionary app. I envision, e.g., instances with a focus on medical terms, instances with a focus on programming/technology, et cetera, but any user can look up any term in any dictionary.

Of course a purely static site doesn't need to federate, but they _can_ take advantage of other decentralization schemes, and news readers should federate or be totally client side.

@tindall "Web scale" sounds like it describes a scaling technique whereby a concept is scaled up through a web of decentralised nodes using web technologies. I have no idea what it actually means beyond it being a buzzword.

@irl "Web scale" means high availability + high throughput: the kind of scale you need to be successful when running a large commercial Web site.

@tindall kinda reminds me of counter-antidisintermediatisation

@tindall
The alternative is "human-scale" software. Check out this article about it:
“Human scale technology" https://medium.com/@jkriss/human-scale-technology-75da763eb03

@tindall not. particularly? there are lots of reasons that, e.g., mastodon.social scales differently then the network as a whole does

@nightpool (just to start off I want to make it clear this was just me jotting down a thought I had and then expanding on it a little bit)

but yeah I agree there, I don't mean that you can just infinitely scale a federated service while appearing to be one instance but rather that the core service can be highly distributed

I mean ideally (from my perspective anyway) m.s wouldn't exist

@nightpool (or rather it would but would not have that many more users than other semilarge instances)

@tindall like I agree that theoretically, mastodon-as-a-network can scale to the size of twitter while still maintaining approximately the same resource usage pre instance.

but in practice there are lots of technical hurdles before we get to that stage. not least of which is content discovery and discoverability.

@nightpool @tindall really hope groups land soon cuz they're gonna be huge for this

@chr @tindall hey look i take what I can get

@nightpool @tindall oh dnang nice

@nightpool Nice!

@nightpool @chr @tindall woooooah

@nightpool for sure. It's not necessarily _easy_ and I absolutely recognize that but if the same resources were put into Mastodon or PeerTube for a few months that are put into Twitter for 2 weeks we would have a much better ecosystem

@tindall i mean, i agree with you. i dedicate a huge chunk of my free time to mastodev. but i want to be precise about where mastodon is realistically at this point in time and what kind of users it makes sense for

@nightpool Yeah! I'm totally with you there. I was speaking in more of an abstract sense

Also thank you for spending the time on masto dev

@varx As I've generally heard it, "web scale" means a combination of very high throughput and low latency, without the loss of the "eventual consistency" property of your datastore.