Who Does That Server Serve? (2010) | Hacker News Reader

Who Does That Server Serve? (2010)

gnu.org

43 pointsredbell24 comments

>Who Does That Server Serve?

What does that server serve is something I wish people kept better track of. Ops people are used to focusing on whatever "big iron" their companies consider as the server for the services they sell, but then people forget "client-server" as a model: There are name servers, web servers, mail servers (SMTPd and POPd and IMAPd), file servers and display servers, etc.

Please people, don't lose sight of this: one box can run all these servers, and if you actually want to be full stack, your box should run all of them. Sure, at scale you need a NOC and multiple web servers pretending to be one web server, etc, but if we lose "server ubiquity" it makes the problem of walled garden solutions so much worse.

at its heart, this is a problem of protocols. Protocols need to be open, creating a space where both clients and servers can flourish. Keeping servers in the raised floor server room is Larry Ellison and Bill Gates dream, but it's our nightmare. With open protocols, it doesn't matter if a particular server is open or closed source, because you will always have a choice. This is the philosophy that created the explosion of the internet. Don't let it die.

It's good to see the GNU project finally recommend AGPL (prohibits use in proprietary services) over GPLv3 (prohibits shipping software on practically all modern consumer devices).

Maybe they could update AGPL and call it GPLv4, so the "or later" clause would kick in and fix things for real.

Doesn't the "or later" clause give the user the option of using the software under the newer license? I'm pretty sure they can keep applying the GPLv3.

Yes, until the original project starts releasing new changes under the new license. From there, the "selfish" party would have to fork and maintain what they have on their own.

That doesn't require GPLv4 though. Anyone wanting to switch their project to the AGPL can just start releasing new versions under it. Existing code would be available as GPLv3 and the newer stuff under AGPL, same as the v4 scenario.

Yeah, when I was planning to switch my personal Apache-licensed stuff to GPL, I decided to switch to AGPL instead. Most of my stuff makes no sense for it to be used as part of a web service etc, but it doesn't hurt to make it future-proof.

Not sure what you mean by this.

GPL s/w runs on all desktop OSs, and at least on Android (see many examples on F-DROID).

Maybe you're refering to iOS? I'm not sure about the collsion between GPL and Apple's terms.

It's completely possible to take a piece of GPLv2 or v3 software, make extensive changes, and provide it only as a web service without ever being required to make those changes available to the community. AGPL solves for that important case, where if you use AGPL software to provide a service and make changes, in addition to providing the service you must also make your changes available to the community.

But I think I misunderstood your point, you were focused on the GPLv3 itself. Which I don't think prohibits running on any computer as you say.

You're correct that I was refering to GPL s/w being able to run on "consumer devices".

But to your first point:

I'm not a lawyer (thankfully), but my interpretation of GPL: that the s/w source must be published if the s/w is "distributed", would require the source of SaaSS s/w to be published, since the "service" is distributing that s/w.

Is that not correct?

> the "service" is distributing that s/w

If it's running on their machine, they're not distributing it; users don't actually receive the software to run themselves and are instead making requests to a server that is running the software. (Client software would be different, of course.) I am also not a lawyer so perhaps take with a grain of salt.

>GNU project finally recommend AGPL

are you sure? AGPL was developed in cooperation with FSF which developed/published the GPL license; GNU is a software project published with the GPL license, but the license and GNU are connected at the hip with Stallman.

The GPL2 license, which GNU was originally published under, did not allow substitutions of new licenses, including GPL3 and AGPL which may by why they might not have appeared "supportive".

Replacing a product you sell, which results in a periodic, unpredictable, and spiky revenue stream, with a subscription or lease is an explicit goal of American business theology.

You as a software developer will never convince the guys that hired you to NOT do that, because it's literally and clearly Good For Business.

Good Software development, good Product development, etc, is Art, ie knowing which tradeoffs will benefit a subset of users, knowing where to spend limited time optimizing for the best results, caring about aesthetics and accessibility and interoperability, caring about software freedoms...

Business Management abhors art because even if you as a CEO or something value it on some level, it's impossible to generate the kind of graphs that please the beancounters from artistic merits.

You won’t convince them because you don’t understand why they think the things they do. Even if you hate management and business types altogether, you’re never going to be able to understand them unless you actually consider they may have a rational basis for their actions. In this case, they don’t care about subscriptions because they hate art or hate freedom or have a “theology” or anything like that.

Subscriptions are preferable to one-off payment for a software product primarily because software incurs an ongoing cost (support, maintenance, enhancement etc which users have come to expect and also servers etc for any online features the software needs) and therefore it’s better to match that with an ongoing revenue stream. If you have an ongoing cost base but only point-in-time revenue then every sale you make is essentially creating future problems because you are creating future cost that isn’t supported by revenue. You essentially have turned your business economics into a Ponzi scheme and can only keep going if you perpetually grow sales which is not sustainable.

Rich Siegel would like a word.

There absolutely is a theology of business, and currently it is this: make as much money as possible.

Commercial software existed long before the subscription model consumed everything, and it was, and still is, sustainable. Subscriptions didn't consume the software ecosystem because the alternative is not sustainable; subscriptions took over because they make more money (and often for less effort).

It's fine if you want to describe that as rational decision-making in business contexts, but I have to object if you cross a line into arguing that there are no viable alternatives. There are subscriptions, and then there are less profitable alternatives.

Exacly. Additionaly, you can sell software and include a year support bundle. If you want to have longer support (or extend support), you have to play extra (pear year). That sounds reasonable.

> also servers etc for any online features the software needs

This is usually a deliberate choice made by software companies. They take their software that could be (or used to be) run on the user's local PC and then unnecessarily move it to a server. Then, they conveniently have a "Well I have to run a server" excuse for changing to a subscription model. You wouldn't have the costs of running a server if you didn't decide to shoehorn that server into the software in the first place!

Even if you only sell shrinkwrap software there are ongoing costs for your business in the form of sales, accounting/payroll, office rent, customer support and so on.

If you want to have bugfixes available for customers you have to pay developers to do that, and infrastructure for releasing them - this is an ongoing cost.

The problem is orthogonal to where software runs: there are subscription models for fully downloadable software that doesn't have to interact with a server (jetbrains for example, and some games).

All of the logic in this entire sub-thread is based on the idea that thet s/w has to be ssold by a vendor.

Large orginazations, representing classes of users (architects, engineers, people who type documents and calculate numbers) could produce s/w for their own use, not to sell, and make this s/w available under a free software GPL license.

They would save the money they would have spent buying this s/w, and spread those savings to everyone else who used the s/w.

I don't want to retype or paste the entry I made below, but please read it and respond to this alternative.

Thanks for the discussion!

The standard way to do this was to sell a license for a specific major version. It goes through a lifecycle where it's initially the most current, then gets superseded by a new major version and maybe gets a few more bug fixes, and at some point one of those releases is the last one for that major version and it's done. Buyers can continue using it forever, but there's no maintenance/enhancement cost, and support can be stopped too if you want.

This worked for decades and continues to work today for some companies. It's not a Ponzi scheme in any sense.

Your statements all seem correct, but its important to point out (and I think this is what you're trying to do) that the logic is based on meeting the needs of vendors/businesses, not the needs of users.

What percentage of people own businesses? What percentage of people use things? ergo What benefits more people: reducing spending for products, or increasing business profits?

Some things are best resolved by a profit seeking model, but far far from all things.

How many billions of dollars in expenditures would be saved, if institutional users (schools, universities, large businesses, trade/professional organizations) developed s/w for their own use, and shared it via a GPL license?

Instead, (statistically) all of these orgs accept the "philanthropic" "gifts" of for-profit s/w products, and subsequently lock all members of each demographic into a lifetime of purchasing that s/w.

It's like giving out "free" heroin in high schools. The lifetime profits are enormous.

Think AutoCAD, but also everyone paying for simple things like Word/Excel/etc. Often, these desktop applications have already been paid for by the h/w maker, making it even more obfuscated from the user.

Widely deployed software freedom would save users billions of dollars, but instead of presenting this as a giant saving, it's presented as a terrible detriment to "business".

This is the propaganda that has brain damaged the US population. The idea that benefiting corporations is benefiting them.

I’m not saying that it’s benefitting users. What I’m saying is if you actually want to make the situation better, rather than high-handed talk about corporate theology and hating art etc and pretending everyone in a position of business authority is either a moron or a zealot or both, you have to meet reality head-on and try to understand why they think the things they do.

If you do that, you can hopefully put yourself in a position to influence them to do things that are more in users’ interests. If you don’t, you may well feel superior, but your ability to make change will be severely hampered.

While I definitely sup from the RMS Kool-Aid, I can’t help but notice that it sounds a lot like small-government libertarianism (s/government/corporations). You’re the product, and all that, but maybe that angle dilutes the point he’s making. Why should software be any different to basically everything else; unless you’re one of those crackpots who bury gold bullion because you don’t trust the man?

Freedom is being able to decide where I get my things done. I can choose to bake bread at home in my own personal oven, and sometimes I do. Or I can choose to buy bread from a bakery. The act of buying bakery made bread is not, in itself, an unethical decision. It's a task that I'm outsourcing to someone else - which is how an economy works. Ethics come in when discussing how the bakery makes their bread vs my expectations. If they're cutting sawdust into their flour, then we might have an issue.

I have the exact example problem mentioned: translation services. In my specific case, I'm helping research a lot of old, handwritten documents in my non-native language. Turns out LLMs are surprisingly good at language and are rather useful for this. I've toyed with running them on my laptop, but the current state of the art is just a bit better and runs faster on someone else's server. In deciding whether to do this job locally or on a SASS, I'm weighing a number of factors. But whether using a SASS, at all, is 'doing me wrong' is not one of them.

I think the principle that would apply is only choosing SaaS solutions which you could, if wanted, run locally (even if it's not a practical consideration at the current moment due to time, power, or space constraints).

Reply on news.ycombinator.com