Hacker News Reader

Comment by pjc50 | Hacker News Reader

Interesting to put this on the front page next to the "immutable infrastructure for mutable systems" one.

The thing is, these drives pull in opposite directions. If it's easy to change it's easy to break, and changes drive entropically downwards as it's always easiest to make the quick and dirty change.

Malleable systems are great for expert users, but can easily be turned into a ball of mud by a non-expert. Hence the popularity of Chromebooks.

Security also enters into consideration. Anything that can be programmed can be attacked. Any program that reads external data is an attack surface.

Look at the history of e.g. web browser extensions, and how they have become more and more locked down to prevent abuse by bad actors.

View on HN

On the other hand, malleable systems are responsible for enormous increases in the productivity of computer users. We may all hate kludgy excel spreadhseets with a nightmare of tangled hacky VBA code, but you see that sort of thing in a lot of businesses' critical workflow because that kinda of malleability by "untrained" users was invaluable in the earlier life of the business.

It's still invaluable in highly locked-down office environments. It lets users automate things that they wouldn't be allowed to automate otherwise (needing a full development environment or the ability to run arbitrary executables or access to Powershell).

I can see immutable infrastructure and this concept working in a copy-on-write model. For instance, if a service is up and running, a modification can consist of duplicating the infrastructure so you have your own copy and then making the modifications. It localizes the change, and in theory keeps the original unmutated.

I have no idea how this would work in practice. This manifesto seems to go against the grain in other ways, as I'm not sure how it can coexist the economic model of SaaS that has emerged. Also, as others have said, the security and UX implications are big concerns that threaten the feasibility of this idea.

And in the intersection of mutability and immutability you have fantastic experiences like live editing and time travel. /Where/ the interface for mutability is located is essential.

I agree there's a tension.

There are strong engineering impetuses for control, for provability, for ahead-of-time decisions, for immutability. Engineers get woken up when things go wrong, they get blamed when things go wrong, & they want, they have to deal with other engineers code doing bad things with their code, & the software world has a wide variety of strategies, tools, & outlooks that all try to build safety, certainty, that restrict & constrain systems, to keep anything unexpected from happening.

I feel, though, that characterizing malleable systems as being for expert users alone is a great injury. At present this may match expectations, but I don't think it's fair to make an assessment of how or what computer-users want when, at the moment, they don't really have the options. There are too few malleable systems out there to judge. What general malleable systems we do have require advanced expertise to use- AppleScript, DCOP (Desktop COmmunication Protocol), DBus, WebExtensions all require at least some programming style skills to use. But I think this expertise is owing to a broad lack of investment in building malleable software. Most software strives to give the user a specified & specific experience, not be a malleable system folks can get the experience they want out of. I don't think we know how much latent pull users have, how strong the tension is to be released from engineer's certainty, because I don't think there are even many testbeds, much less a broader societal view of systems flexibility, systems adaptability, of softer software.

To me, I just think it's clear that, whatever valid concerns we do have, whatever justification we have for building endless arrays of rigid systems, it seems like it's not worth it anymore. We keep making more and more software, the profusion of software applications is a torrent of similar-but-different applications, focusing on shuffling this or that kind of data specifically, yet the generality, the inter-connectedness of the pieces seems so weak, and it seems like we've shut users out from having any creative capabilities & control of their own.

I think WebExtensions are a very interesting case. The killer aspect, to me, is that the extensions work invisibly. We can't rely on everyone becoming an expert extension operator, but I do think that some reporting mechanisms on what extensions do would radically reshape how the community review's & socialized the extensions they use. In many ways, it's that the extensions are a means of exercising the malleability of hypertext, but they themselves are rigid & unmalleable & opaque. A part of being malleable- a precondition- is making yourself visible, providing viewability into what is happening, such that folks can reshape & redefine & redirect those actions that you the software are enacting. So the circle needs to grow more, is what it seems to me- rather than lock down extensions, rather than rely on authoritarian centralized powers to moderate extensions, my plea would be to make extensions malleable too, to make their working visible too, and to try to let consciousness & human adaptability start to take hold in this realm.

You're absolutely correct that there are good reasons why software is so rigid & inflexible & opaque. Security has been nightmarishly cumbersome to the whole industry for the last decade. And I think malleability does indeed open up a lot more surfaces, endlessly opens surfaces. But at some point, we have to develop systems that have a co-relation with their operators, that respect the operators, that let them see what is happening & let them work their own systems. Continuing to try to push computers forwards entirely on the assumption that the user is worthless, does not know anything, and can not understand what is happening has cornered us, has cut off the chance to try & explore other options.

Reply on news.ycombinator.com