Open Source is losing, SaaS is leading, APIs will win… | Hacker News Reader

Open Source is losing, SaaS is leading, APIs will win…

medium.com

42 pointsvr369051 comments

Not quite.

If I want to start a project in, say R, I don't think I will start by searching "Open Source". I will search for "R". Here is the trend: https://www.google.com/trends/explore#q=software%20as%20a%20...

So, this analysis is not accurate. Someone searching for "Open Source" might have been prevalent when the OSS movement was nascent. Now I think people will look for specific technologies rather than blanket terms.

Plus, the consensus has been reached by now, I'd imagine, that github, bitbucket, and gitlab are the best places to go for open source software. If you just wanted to explore the world of open source, you'd either search one of those terms, or go directly to the site.

I use "open source" when searching for Windows applications, to avoid the trialware etc

Yes, this is the fallacy of using search terms as the data for any analysis that isn't about search terms.

Yeah, I always cringe when people claim to have made some huge insight based on Google search trends.

Yeah, you have to be careful which conclusions you reach from these trend plots. By the author's arguments, I could argue that maybe people just won't be interested in software at all anymore in a few years: https://www.google.com/trends/explore#q=software%20as%20a%20...

A lot of these concepts are just too general to search for. That doesn't mean they're less popular, it just means their conceptual boundaries are too big for a single phrase.

> Today, the CTO of a company can spend engineering resources to build their search solution with Elasticsearch or buy speed with Algolia’s SaaS.

I think OSS has the long term advantage here.

Everyone needs log storage and search. The combination of OSS and highly competitive IaaS will drive this part of the stack towards commodity pricing and little total cost of ownership (setup, and operations)

It's already happening. Elasticsearch is quality software. You can provision an Elasticsearch cluster on AWS with a click of a button. You can also store all your logs in CloudWatch forever.

This stack costs pennies to run compared to logging services like Splunk.

Yes configuring it takes some time but an end state for this container generation is much less config.

Now you own your logs and data and search. Not some SaaS vendor.

Disclaimer. I'm working on an OSS platform, Convox, with the goal of helping the CTO get cost effective and reliable private infrastructure without big setup burdens and huge SaaS bills.

I think it just depends on whether you can afford an ops person. If you have one, setting up your own stack makes financial sense, as it'll be cheaper. If you don't have enough ops to justify hiring one, then you might as well pay for the (more expensive per unit) SaaS offerings that will be cheaper purely because they're still only a few services.

That assumes the SaaS option doesn't also require some amount of ops. An assumption I wouldn't necessarily bet on when you start using SaaS from multiple providers to satisfy your various needs.

Sure, but SaaS generally require less ops. That's what you're paying for, in the end.

That's what you're most often paying for now, but that's not a fundamental property of SaaS. It's a byproduct of the fact that most are built on open source products. If proprietary software companies like Oracle and Microsoft went purely SaaS for their business models (thinking of DB software here), you'd be paying just to get access to the software, in addition to any ops overhead benefits you also gained.

Does Algolia have any particularly awesome client stories?

https://www.algolia.com/customers

The HackerNews search stands out to me as surprisingly weak in terms of intuitiveness, and oddly-branded to boot. A lot of their other clients seem to be hype-y startups.

The customer-list for ElasticSearch seems a lot more robust and battle-tested:

https://www.elastic.co/use-cases

We have a ton of other awesome (from well-known) clients that are under NDA. Rest assured: we're working with fantastic people. :)

Personally, I love our big clients, but one of the things I love most is that we provide a search that scales from concept on up. While a lot of the customers we've got listed on that page are start-ups, I think it's just as cool that Chrome Industries or Ritz use us for their search. I love that we're able to help these companies that I shop in "IRL" have awesome search without having to hire a specialist just for that.

The big disadvantage of OSS versus (competent) infrastructure SaaS, is that open source software tends to have very poor operational efficiency. A major component of the infrastructure SaaS business model and defensibility is arbitraging the OpEx between open source and a properly designed SaaS implementation, primarily by radically reducing resource requirements to deliver a given level of scale and performance.

The rule of thumb for most infrastructure SaaS like databases, big data platforms, etc is that you can get a 10x reduction in hardware requirements versus OSS, hence the OpEx/CapEx savings, with a well-engineered implementation of the same functionality. As someone in the business of doing exactly this, I can tell you that this is simple to achieve. OSS may be "free" but if it requires an order of magnitude more hardware to deliver the same performance and scale as my cloud implementation of the same thing, it makes the OSS much more expensive to the end user.

This dynamic would go away if OSS infrastructure software was designed with performance and efficiency as a priority, but in practice that has been pretty far down the list of priorities for most popular OSS software. But I think most people are surprised by just how inefficient OSS software actually is compared to other implementations. I have been telling people for some time that without an increased focus on absolute performance and efficiency in OSS, it will be increasingly marginalized in usage because the CapEx/OpEx of deploying it is far too high compared to the alternatives.

Could you provide some high-profile examples where the SaaS implementation is much more performant than the OSS variant? Genuinely curious... I figured the SaaS vendors (mostly) use OSS under-the-hood anyways, and then they have to take their overhead and profits, etc.

Arent most SaaS backed by the same open-source software? Often the part you're paying for is the management of the software, not the software itself.

This is a really great perspective.

You're totally right, SaaS companies have far more motivation to maximize performance and operations to reduce the cost of keeping their services running.

OSS often doesn't have the need or the expertise to do this.

But I do see some of this shifting...

Some OSS projects are now built by folks that have lots of performance expertise and goals.

There's also more funding or sponsorship for OSS development which will help.

For example Redis is ridiculously cheap and easy to run on your own.

antirez is a world class engineer with high standards and he has sponsorship to work towards his goals full time.

AWS offers it as a service through ElasticCache which removes what challenges remain about configuration, replication or failover.

So I wouldn't build, integrate or bet on a SaaS "cache" platform/company any more. It's been commoditized.

I see signs of the same thing happening to more parts of the stack. ELK isn't the most performant log analysis platform, but it's close to Splunk in my experience, with the added benefit of the possibility of anyone, Elastic or the community, contributing performance improvements.

Yes there is lots of wishful thinking here...

This smacks of a complete misunderstanding of SaaS and FLOSS. How did this person even get into software development? Seriously, because to me if a SaaS is using a FLOSS licensed library and regularly contributes to it on its github repo then it's clearly not a matter of FLOSS in decline. It's just that FLOSS as a whole has become the ordinary of software development (just like CAD/CAM is the ordinary of aircraft manufacturing).

Also, using Google Trends to prove a point on something that requires you to see actual participation just proves my skepticism is right on this point. Google Trends doesn't look at git commits. Google Trends doesn't look at symposiums or standards organization contributions (cash, research, code, etc). All it does is look at search words. Wow! You found out that people aren't searching that much on the subject anymore, WTG! /s

This whole article is predicated on the wrong measurement.

When you search Google Trends for Open Source Software as a type of Software Industry [1], you get a much different picture.

[1] https://www.google.com/trends/explore#q=%2Fm%2F01pjyj&cmpt=q...

At this point in the game, Googling for Open Source is like talking about oxygen. I don't even think about it. I just breathe.

100% agreed, but for different reasons. I recently realized that I've been mistaking google trends as accurate feedback on the popularity of a subject. It is not.

My current thought is that it much better indicates a lack of consensus (or foreknowledge) of the proper source for information about a search term. If, for example, you know that godoc.org is the best place to search for golang libraries, you'd probably just type godoc.org. If not, you'd probably just google "golang rest library" or some such term, adding to the results for golang.

Once those sources become well-known and begin being passed around on side-channels (outside the reach of google searches), then google trends is no longer accurate. For any given suject, it's impossible to get a gauge on how much traffic is just going directly to bookmarks or links.

Unless you're collecting anonymous usage statistics from a market-leading browser. For example. :-D

This article feels like it's saying "measurements show that the sky is actually yellow!". No, it's not. I can plainly see that open-source code is 99.999% of my stack, with a tiny bit of proprietary code (mine, and even much of that is open-sourced by me in library form) on top. If your measurements show that OSS is losing, your measurements must be wrong.

anecdotes != data;

That's another fallacy. Anecdotes are data, they're just an extremely weak form. A million anecdotes may be better than a single data point.

I disagree. While SaaS blinks in and out, requires maintenance, and most fading away over time when interest or business model dwindles, open source projects lives on even in orphaned form. Time is on open source's side.

If you add a trendline for the term 'Software' you will see that it has been steadily decreasing at the same rate as 'open source'.

Came across this article on Medium and wondered the people here on HN think about it.

I am not sure about the author's methodology for comparing OSS with SaaS APIs. Looks very pseudo-sceincey. But if there's any truth in the conclusion (that people are leaning more towards SaaS APIs than OSS), I am worried over time it will increase the barrier to entry for building new software thus resulting in fewer awesome stuff unless you have deeper pockets.

The three trend graphs at the tops are disingenuous. Here they are in relative comparison to one another: https://www.google.com/trends/explore#q=software%20as%20a%20...

Software as a service is a completely flat line by comparison.

adding "saas" makes a bit more of a difference: https://www.google.com/trends/explore#q=software%20as%20a%20...

I did the same thing just before you replied. The difference with "saas" is super trivial though.

Another thing: Look at the geographical distribution for "api" and observe a huge peak for Indonesia, stomping everything else into the ground. I'd guess that "api" is some sort of very common word in the Indonesian language.

I don't agree that the free software business models are failures. Many large companies need actual enterprise support and licensing to fulfill their business requirements. Just because most start-ups (which is clearly the author's circle of friends) aren't big enough to need such support doesn't mean that it's "failed". Support is a real value add, because it means that you have developers on call that can maintain the system you based your stack on. The fact that the code is free software doesn't make a practical difference to you (aside from moral differences I won't go into here).

> Open Source is losing, SaaS is leading

And what are these SaaS apps mostly built with?

Open source software. They usually build on a mix of technologies that can be called an "open source web technology stack".

I even believe the current effort of M$ to open source some of it's technologies is geared at making sure those do not become irrelevant in the SaaS-age.

Eh, the search term Open Source is becoming less frequent simply because it has become the default and software that is not Open Source has become the exotic case. Just search on google trends for 'github' which has basically become a substitute word for 'open source' and you see a nicely growing curve.

Maybe one of the reasons people don't search for "open source" as much as they used to is that it doesn't need explaining any more. Anybody who has reason to care has already been exposed to the concept by now.

> The increase in development speed is driven by the innovation of re-using existing blocks. Basically, developers don’t need to reinvent the wheel every day.

In the 60's, when TTL integrated logic chips appeared, the digital hardware world took off. Anyone could design at a higher level of abstraction. For many years people said that there needs to be a TTL of software. It seems we might be getting there.

It rather seems to me that software that has been successfully established feature-wise as service will eventually get copied as a off-the-shelf OSS software (see the discussion about cloning Slack)? Basically like commoditization of other goods, but in the information economy with zero marginal costs obviously being a free commodity.

There's some good points in this article, but those are lost in the false comparison of OSS/SaaS/APIs.

That's not even like comparing apples to oranges, it's like comparing apples to apple trees.

What does Microsoft do these days? Open source many of its software! And OSS is preferred not because it's free & open, because it works and it gets things done!

> What does Microsoft do these days? Open source many of its software! And OSS is preferred not because it's free & open, because it works and it gets things done!

I prefer free software because of the fact it respects my freedom. Even if there was a "better" proprietary editor then Vim I wouldn't use it. Stop spreading the wrong message, practical convenience is a side-effect of freedom allowing innovation.

"All this has happened before. All this will happen again." -- the writings of Pythia.

This is why digital humanities education is important and useful.

It stops this kind of mistake.

I find it quite annoying that recently web hipstors apparently hijacked the term "API", and now everyone assume that by default "API" means something totally different from what it used to.

What does "API" mean to you (or originally meant) vs the hijacked term?

Wikipedia have not succumbed to the hipstors so far: https://en.wikipedia.org/wiki/Application_programming_interf...

API is an exposed interface of any library, standalone program, etc. It's a very wide term. But stupid hipstors reduced it to their stupid understanding of a web RPC.

You are not wrong, but your tone is very rude.

Yes, API means much more than web RPC. But the "web hipstors" are not wrong in calling their services API either. Educating people about the multiple other meanings of that word is fine, condemning people that use it for a limited (but correct) meaning seems very unnecessary to me.

My tone is rightfully rude. I hate what the hipstors are turning this world into.

It is somewhat annoying when I mention that React has a "slim API" and people ask for an http endpoint...

I guess the reason is that web service don't sound as glamorous.

"Hipstors" didn't hijack shit and "API" is correct usage here. The fault is with the (hypothetical?) morons who think "API" only refers to web services.

When was the last time when somebody used the words "web API"? When web monkeys say "API" they expect anyone else to understand it as the new default hipstor meaning. It is extremely annoying.

Reply on news.ycombinator.com