A Proposal for an Antifragile Software Manifesto
sciencedirect.com"We value reliable software", is nice, but everyone values that.
But if you said "We value reliable software over meeting deadlines", then you'd actually have said something.
The power of the Agile Manifesto was that it clearly identified tradeoffs they were willing to make.
> if you said "We value reliable software over meeting deadlines", then you'd actually have said something.
This reminds me of a Jamie Zawinski quote from the book "Coders at Work":
> [...] Are you trying to write good software or are you > trying to be done by next week? You can't do both. One > of the jokes we made at Netscape a lot was, "We're > absolutely 100 percent committed to quality. We're going > to ship the highest-quality product we can on March 31st."They try to prioritize their principles, but they are vague. Let's see:
> Our highest priority is to satisfy the customer by building a non-linear, proactive, and self adaptive system
Okay, firstly, what is that? And secondly, does it serve the customer if those are not the customer's requirements?
What if I don't want a "self adaptive" system, whatever that is? It could be bad: self-adaptive software keeps some global state somewhere and doesn't quite ever do the same thing twice. If I require something that produces exactly the same output every time I feed it a given input, then I'm ill-served as a customer if I'm handed some self-adaptive contraption.
> what is that?
A collection of popular buzz-words in contemporary software engineering research.
> And secondly, does it serve the customer if those are not the customer's requirements?
These are never any customer's requirements. Not a sane customer, at least. Instead, they are (conjectured, probably correctly, to be) generalizations of sufficiently many customers' requirements.
But you're correct -- most customers don't want or need any of those things (except non-linearity. But every piece of software everywhere has this spades.)
I agree they seem like buzz-words, but not from software engineering research. They look entreprisey or startup friendly buzz-words to me.
Or you could read the book. I haven't (summer reading list, again). But it's got a 20-page bibliography, and appendices with all sorts of greek letters that define, well, a lot. I think.
The Black Swan was also by Taleb and probably the most influential finance book of the last 20 years, plus a great read for anybody thinking in 'systems'. It'd give the idea the benefit of the doubt. Sometimes words have no meaning, sometimes it takes some work.
The parent didn't ask what antifragile means. The parent asked what non-linear, proactive, and self adaptive means, especially when taken as requirements for a software system.
Reading Taleb's book to answer that question is probably a lot like trying to extract evolved dogma out of a religious text because each of those words has taken on a life and meaning of their own in SE literature. You can probably eventually get close, but you're better off just reading a few SE papers on those topics and inferring the authors' intended meaning.
A URL to a git repo showing the non-linear, self-adaptive piece of software would be more effective.
Antifragile refers to a property in which a system does not just resist breaking down under disorder and stressors (robust), but gains/learns/becomes stronger. A good example is how bones heal back stronger where they were broken.
While we are currently able to build self-healing systems of a sort, fault-tolerant Erlang systems being a good example, it doesn't seem we can build systems that are truly antifragile on their own, that would require strong AI. Otherwise, humans monitoring and intervention are required to improve systems under duress.
This manifesto does include "team" and "organization" to account for that, but a true autonomous antifragile system is a ways off.
Software such as fail2ban that bans harmful IPs and spam filters that learn over time would qualify as antifragile. I can't think of anything at the system security level.
Your bone example isn't accurate (http://www.nytimes.com/2010/10/19/health/19really.html). The example used in the book are muscles (stress -> breakdown -> creating stronger muscles)
Maybe more accurately, bones subject to survivable progressively more intense stresses are stronger. It would be like an autoscaling algorithm under increasing loads (not delta function loads).
I think a better example is balancing pendulums. They stay upright under random movement, but fall when things go still...
TL;DR:
1. Our highest priority is to satisfy the customer by building a non-linear, proactive, and self adaptive system.
2. We welcome changing scenarios where unexpected events (Black Swans) are the real paradigm shifting entities.
3. We deliver assuring embedded and adaptive fault tolerance.
4. All stakeholders, and the broader environment, lead the antifragile organization.
5. Build antifragile projects around motivated, skilled and open minded people. Give them the environment and support they need, and trust them to get the job done.
6. The most efficient and effective method of building an antifragile organization is building on honest, open and transparent communication.
7. Continuous exposure to faults and automatic fixing is the primary measure.
8. An antifragile organization promotes a context aware environment. The stakeholders should be able to maintain a system indefinitely.
9. Continuous attention to technical excellence, reality, redundancy.
10. Error loving - the art of learning to be antifragile – is essential.
11. Antifragile architectures emerge from self – organizing, context aware teams.
12. At regular intervals, the developing team reflects about the context situation, on how to become more effective, then tunes and adjusts its behavior accordingly.
Why do people still keep writing Manifestos in 2016? I thought it was evident they don't work out - or that they get "abused" as political opinions.
We don't need Manifestos, we need people who think about each situation and problem differently, basing their thoughts on experience, knowledge, and communication. You can write all the s* you want, but you won't get that thanks to 10 lines - which will be always ambiguous and misinterpreted, then what? Will you write another manifesto?
We need experience, we need to learn, we need to improve. You literally need to fail to get any improvement. And companies need to keep their people, they need to invest in people, instead of relying on a bunch of buzzwords or manifestos to prove themselves "cool" and up-to-date. Stop doing that. Make a step back, give your people time to think, talk to them, and give them the right resources.
We may need those things, but it's increasingly clear that we cannot afford them so far as we know. I'm using "afford" widely - if we needed those things and understood that need properly, we'd have them. It's the mythic "$20 bill on the ground" - if it was really there, someone would have picked it up by now.
it's no great secret how to write robust and unfragile software. We apparently only lack the will to do it.
I've been in spots where we priced a defect, and they still didn't want it fixed.
I have also worked for such companies. It's more common than what you and I think :)
However, they cannot afford them because they simply don't want to be the best. Look at those companies who invest in people and give them time to grow/think/be. Their products are above average. Why?
When you want to be the best, you simply have no other way than "thinking" and "reasoning". Unless you are just lucky and you get one or two Einstein[s]. But even then, if you don't give such a genius the right environment, he is gonna do his best to get away from you.
What you call will I call it attitude towards business. Nowadays we don't write software anymore to make a product last at least 10 years, we build companies to be sold in order to get someone (the founder) rich so that he can invest in new crap to sell. Why should he waste 500-1000 $ when that money can go to a new feature or an "important" bug (coming from a bigger customer)?
I think a manifesto with a lot of buzzwords ends up being quite fragile.
I suspect you might not be applying the right ontology to fully elaborate the reification of the concept.
They address that in the article. This is a spitball proposal.
As with any manifesto, words, interpretation, and practice are where they fail. I've been through the Agile/Scrum classes, yeah, great. Does management buy in, despite professing such? Yeah, not so great.
Unfortunately, for me, whenever I see the word "manifesto" in relation to computers, this is the one I immediately remember...
And this is the one I remember instead:
After reading that, the world is now cyan. Thanks.
This will be difficult to do in text (imagine a triangle):
Quality
Time Cost
Honestly, I would love it if anyone could make the decision at all. Usually it's "I want all three".
Yep, managers seem to think the triangle can be overcome with enough money/prowess/brute force.
Well, if they're willing to invest money, then they've chosen, no? Cost is not important, but timeliness and quality are.
Cost is always important, everyone has the price at which they're no longer comfortable.
Money in many cases can buy you (some) time.
It can't be quality because "[e]rrors are the primary source of an antifragile system", quoting from the paper directly.
The triangle is "cheap", "fast", and "highly available".
Pick any two...
You can have all three. Prioritizing quality from the start, by doing design reviews, pair programming, tdd and so on, will actually reduce time needed to deliver fully functional software, and therefore lower overall cost.
reduce time needed to deliver fully functional software
Except you can deliver a cheaper product faster if it's not fully functional, hence the triangle.
In theory.
In practice? Pick a spot inside the triangle.
Interesting that the concept of Antifragility is gaining traction with regards to software.
Back in 2014 I wrote a blog post "Antifragile Software Ecosystems" that discusses how IMO antifragility relates to how software is developed.
https://opkode.com/blog/2014/01/14/antifragile-software-ecos...
I was kind of hoping for a document that outlined principles to make the engineering process truly antifragile, i.e. one that would thrive under contingencies.
Honestly, I have trouble fathoming how such a thing would be even possible in the realm of software. Maybe a system that "learns" previous failures and proactively monitor failure conditions / throttles contended ressources? (Seems quite specific already)
And the document surely didn't tell me.
I actually think the former would be more within in the spirit of Taleb's book. software is the result of a complex, sometimes problematic process and not every piece of software needs to be antifragile. when bugs or system outages pop up it isn't the responsibility of the software to fix itself, it's the responsibility of the team behind it.
Chaos engineering falls in this category: http://principlesofchaos.org/
I think this is more about tools than process. Specifically tools to assess fragility (bad dependency patterns). You cam only manage what you can measure.
When in 2009 I was calling it "Fragile," not "Agile," people were laughing at me as yet again I wasn't politically correct!
I don't think it's mainly about agile. There is plenty of good examples of fragile OSS out there that has nothing to do with agile. Entire npm ecosystem is very good example.
It's not only about Agile, but incremental development is always detrimental especially for complex projects. Also, most companies apply some ghetto version of Scrum, most don't even know what SOS (Scum of Scrums) is, pick and treat the Scrum Master incorrectly, don't even do Planning Poker, don't use Epics at all, don't define stories right, don't use burndown charts at all, etc. In most cases, I see Scrum as an eyewash for the business. It's a pseudo-scientific approach to prove to business we've done something this week, and a solid excuse for missing deadlines, etc.
It's not only about Agile, but incremental development is always detrimental especially for complex projects
I really have to dispute that.
First, it's not really clear what you mean by "detrimental" - is that to software quality, or development schedule, or suitability? Or all of them, or others? Agile is very clear about the trade-offs - working products being more important than documentation for example, or responding to change over following plans.
Second, I don't think it's true. I've directly observed, in multiple organisations, that agile development is perfectly capable of delivering good systems on-budget and on-schedule. If anything, it's especially good for complex systems, which are those most likely to require flexibility and have hidden requirements.
I also dispute the idea that "ghetto Scrum" is invalid. Development processes should be tailored to the requirements of the product, organisation and team. Not every team needs epics or planning poker or scrum of scrums, and it's totally valid to take some of the ideas of scrum and apply them to your own development process without having to buy the whole thing.
I consistently see agile development hand waved away with this argument that it doesn't work because there are companies which do it badly, and it's baffling to me.
You can't build a house from the roof, but that's what Agile indirectly makes you do. Although Waterfall is the other dangerous extreme, it makes you think, correctly estimate, and plan. You can't apply Scrum to every project - especially those with firm deadlines! Yeah, it's good for startups, where feedback and market shape your product, by can you build a spaceship using Scrum? Honestly, as an engineer, even when I use Scrum, I still use Gantt charts, and that's another thing that makes people make funny faces in Scrum.
You can't build a house from the roof, but that's what Agile indirectly makes you do.
And you can't compare software with a house, or even a spaceship :) For one thing, we have far fewer constraints.
When we build software we are (generally, though not in every case) building a solution to a problem rather than a strictly defined product - and in many cases, we are unsure about what the final product will look like! Spotify did a good illustration of this process in this slide: http://www.chrisyin.com/images/spotify-mvp.png - the point is to continuously deliver features to customers and using the feedback generated to inform ongoing development, which like you point out, is absolutely essential in some cases.
, but that's what Agile indirectly makes you do. Although Waterfall is the other dangerous extreme, it makes you think, correctly estimate, and plan.
That's perhaps the idealistic view, but there is a cost there too - change becomes much harder. I also suspect that there is no such thing as a correct estimate!
You can't apply Scrum to every project - especially those with firm deadlines!
Maybe not, and Scrum is not the ideal methodology in all cases - but it's also not the only way to be agile. It's also a fallacy that planning in advance means a project will meet a deadline - let's remember that Scrum doesn't typically mean you are doing different work, just that you are slicing it and delivering it differently. I'd argue that one of the main differences is that if a Scrum-developed project misses it's deadline, you still have a working product - it just doesn't necessarily have all the features you'd like. But if you are correctly building the most important and complex features first, then that's potentially much better than the "waterfall" outcome of "the product isn't complete and can't be used".
can you build a spaceship using Scrum?
I probably wouldn't want to try it, but this is an example of a project with incredibly tight constraints on what can be built - far more than software typically is. But I'd imagine it's also possible to build components of any product in an agile manner.
I still use Gantt charts, and that's another thing that makes people make funny faces in Scrum.
That's great, and it's something that should be encouraged - development methodologies should be determined by what works well for you as a developer, team, or company! A Gannt chart, assuming you are changing it as product requirements change, is basically the same as a Scrum burndown chart with additional planning thrown in. And we do the same sort of thing - "these are the epics, this is how long we estimate they will each take, and here is the order we will do them in." It's not so different.
Ultimately these methodologies are designed to help projects cope with change, which has been a long-term problem in correctly sizing and executing software projects. Maybe they won't always work for every project with particular constraints, but I reckon they are far more applicable than some people believe.
