Danil A.

[Danil A. @ag_danil]

@ChShersh @kamikaz1_k Don't you find it absurd? I mean they took a system that forces ordering on a transport level (log) and just fighting it to make it work like unordered system (queue).

[Danil A. @ag_danil]

@Hillgod @ChShersh @kamikaz1_k Are you saying that https://t.co/AjXNNMdH6R is not ready for wide use?

Tbh I am anti-kafka and prefer unordered systems like AWS SQS (and overall think that ordered systems like Kafka is a mistake). But this KIP sounds exactly like they realize a mistake they made.

[Danil A. @ag_danil]

@oliver_io_ @ChShersh @Shreyassanthu77 It is "sync" in a sense that you can't proceed until you get a response. If you are handling a user request and going to other service for some portion of the data, then you can't fullfil original user request until other service responded to you. As so you are blocked.

[Danil A. @ag_danil]

@ChShersh Can't agree more. Finally, I found somebody who is also thinking that way about microservices.

I am advocating for this at our company for many years. But it is hard. People resist a lot.

I had to write a mini-book for all engineers, but even then people keep arguing

[Danil A. @ag_danil]

@wesbos Tailscale on a home router with subnet routing. Split DNS in Tailscale to use my router to resolve domains in *.home.danil.codes. OPNsense + Unbound DNS on it. Letsencrypt HTTPS certs for all internal home domains with DNS validation via CloudFlare API.

[Danil A. @ag_danil]

@softwarevlogger Но в 99 процентах случаев вам нужна асинхронная коммуникации между аккаунтами. А это сразу EventBridge. То есть никаких VPC.

[Danil A. @ag_danil]

@softwarevlogger multi-account это единственный способ в больших организациях. Если у вас больше 3 команд/сервисов, то у вас просто нет вариантов. Плюс аккаунт на environment. (Dev + staging + prod) * 3 services = 9 аккаунтов.

[Danil A. @ag_danil]

@0xlelouch_ This makes me mad every single time. Repeat after me: I. don't. need. the ordering. at. the transport. layer. (in 99.9% of cases).

Just design your application properly. Be idempotent. Ditch sending deltas. Send state (snapshot).

Kafka must die.

[Danil A. @ag_danil]

@Nekrolm Это, конечно, всё хреново, но это имеет мало отношения к js. Исходники почти всегда превращаются в исполняемые артефакты перед дистрибьюцией. Так что я предлагаю думать об этом минифицированном js как о бинарнике. Были попытки глубже эксплуатировать эту идею: гуглится как BinAST

[Danil A. @ag_danil]

@brankopetric00 2nd thing: 200ms is avoidable. If you switch to async communication. Instead of making one service call another in "just-in-time" style, get what you need from it ahead of time - listen for events that carry the data and save copy of it locally (part that matter for the consumer)

[Danil A. @ag_danil]

@brankopetric00 If you try to make all 100+ of people to work in one codebase, it would be a disaster. It was hard to make even 20 people to agree on stuff (language, architecture, SDLC, etc), so instead of forcing people to work in larger group, just set them free (give them autonomy).

[Danil A. @ag_danil]

@thepanta82 @peturgeorgievv @housecor UUID v7

[Danil A. @ag_danil]

@ksuunja Похож?

[Danil A. @ag_danil]

@danilop This is years behind https://t.co/0fB46KAQam

[Danil A. @ag_danil]

@mattpocockuk I am wondering if anything has changed since then. Since then I stumbled upon this issue at least 3 times. And I still can't believe there is no solution. Seems like a really basic feature and there should be some workaround. Maybe I can reorganize code in some other way?

[Danil A. @ag_danil]

I need a help. Calling for TypeScript guru @mattpocockuk.

https://t.co/P0328vl6pO

[Danil A. @ag_danil]

@mattpocockuk I know you specifically mentioned "within Node", but does it make sense to offload that concern to separate machines/cluster? (If we are talking about an architecture of possible commercial SaaS)

Firecracker is the best option for that. https://t.co/VVqG2ZwCCx

[Danil A. @ag_danil]

@MrsPunishka @alinkerka В комбинации с такой реакцией, как тут у многих в тренде, это точно говорит о незрелости (эмоционально стабильные и зрелые люди не будут устраивать истерику по поводу непреодолимой непредсказуемости жизни).

[Danil A. @ag_danil]

@MrsPunishka @alinkerka Если для человека критично опоздание другого человека и это собьёт его "планы", то это говорит что этот человек плох в планировании (напихал в расписание слишком много вещей не учтя хаотичность реальности).