Hacker News Reader

Comment by SubuSS | Hacker News Reader

We shouldn't fetishize control as much.

In >15 years what I have seen is that OSS comes with a huge cost of attention - that you'd be better off focusing on your business need.

Running most of these OSS even at a medium employee count scale is a full-team-commit-nightmare. It essentially means your it department now needs a specialist who will be able to handle your org quirks, remember the tweaks and keep baby sitting it. Not to mention many OSS teams (I don't know about the above listed) are ever so sneakily introducing enterprise clauses about usage in their EULA.

I don't want to name names, but even with paid software we are seeing this problem everywhere in trying to host: But at least here, we got the actual developers and experts on a call who can walk us through why x employees are fine and x+1 nukes their system.

View on HN

We don't need OSS in this case, we need open protocols so anyone can use any other client, proprietary or not.

We've had a perfectly good open chat protocol since... gosh, I don't even know. It's called IRC and it's so simple you can implement a server or client in python in under an hour.

This is what happens when you over-engineer things.

IRC has many drawbacks not shared by more modern protocols. Slack works when you're offline, saves history, has notifications, search, etc.

IRC is not perfectly good or it would still be used.

> Slack works when you're offline

IRC doesn't do that, true, but there's a bunch of ways around it.

> saves history, has notifications, search

Trivial to implement in the client.

> IRC is not perfectly good or it would still be used

IRC is still used. Almost everything any fancier messaging protocol does can, and has, been implemented as a client-side feature with little hassle.

So corporate users can set up an IRC server; a network of per-user bouncers for offline state maintenance; an internal file hosting network for uploads; a bunch of server-based logging stuff and a central search repository to check history; train every middle manager to be an IRC server admin in order to add and remove people from channels for new projects; somehow intervene to disable DCC and other methods users could use to circumvent server-side logging; maintain their own fork of an IRC client in order to implement inline code, images, website previews, etc; break IRC's protocol in their client and server to allow text in excess of the maximum line length; and separately find a SAAS video/audio chat vendor or roll their own...

... all to save $7/month/person (or less for some competitors like MS Teams) because Slack is, like, too proprietary.

They could easily pay someone else to handle all that stuff just like we do with HTTP servers. We're not talking about the service, we're talking about the protocol.

> maintain their own fork of an IRC client in order to implement inline code, images, website previews, etc

These already exist. No need to maintain a fork unless you want to add features that don't exist in it. And since IRC is so trivial a protocol, that's not even hard.

> break IRC's protocol in their client and server to allow text in excess of the maximum line length

You don't have to break shit. Just split lines into multiple messages and piece together messages from the same user when they were also the last one who said anything, or use a character that means "message continues", or any of probably a dozen other ways you could implement that on top of the already existing protocol and that will still gracefully degrade on clients that don't support it.

Every time someone on HN says "you don't need x, just do a, then b, then c and you'll get something similar" I'm reminded of this comment: https://news.ycombinator.com/item?id=9224

Dropbox is just a hosted rsync server really [0]. The service has value, the technology didn't really require a new protocol except for lock in purposes.

I'm not saying that a service that takes existing ideas and makes them easier for people to use is bad, I'm saying that the open protocol people are calling for already exists, but because they have tech-brain they'd rather build a completely new protocol because it's more fun.

[0] Not literally, I have no idea what it uses.

I dug around many years ago into dropbox. I actually found rsync was being used in at least some of its functionality. I doubt that remains the case, but it made me smile when I found it as it made a lot of sense at the time.

Edit: spelling

Slack just freaks out when I'm offline (or it even thinks I am). To the point I often just end up writing an email as it's less painful than using slack.

Open protocols without OSS tend not to be particularly robust. In this instance, we have both: open protocols like Matrix.org, XMPP & IRC - and open source servers & clients (like synapse & Riot.im on the Matrix side), which are no longer an opensource curiousity but ready for primetime as Slack/Discord replacements. Just as Apache is ready for primetime as an IIS replacement.

Yeah, it's worked out well for email.

You can use paid services, or not. It's up to you, but they all interoperate.

They certainly DON'T interoperate. I have a mail-tester score of 100, use SPF, DKIM and DMARC, yet Microsoft still sends all my emails in the spam folder. There's nothing I can do about it, it's a black box. The only suggestion from them is that I subscribe to some paid program with another company in order to maybe, or maybe not, increase the trust in my mail server.

But isn't this not really and argument against FLOSS in this case? Isn't support the entire business model of, say RedHat?

I don't know about the rest of the software but Mattermost is easy to administer and requires minimal babysitting. I have it deployed with a simple docker-compose file which I only touch when I want to upgrade it, and there are Helm charts if you want a more production grade deployment. If I were to deploy it for my company I'd probably use the helm chart + a cloud hosted Postgres as the database and would likely never have to touch it between upgrades.

But you don’t want to go full ducking SaaS with things as well, where you don’t know what is coming until the ‘improvement’ is foisted upon you. Software gone mad. Spoken as a SaaS vendor

> In >15 years what I have seen is that OSS comes with a huge cost of attention - that you'd be better off focusing on your business need.

That’s not even wrong because nowadays almost all software comes with some OSS contained. So the non oss set is basically empty. I wonder if HN comes with a huge cost of attention because it runs on some OSS lisp?!

Reply on news.ycombinator.com