This page is inspired by thinking about PlanetMath, literate programming, WebMathematica, the recent discussion about Bibliographies and Amazon referrals – and about whether it is cool to get funding from places like Google, Amazon, etc. (also about whether it is cool to use their no-cost services), as well as thoughts about free software business.
The basic thesis of my argument is that non-free webservices are not different in a substantial way from non-free services used locally.
If you think that it is OK to use non-free software locally, then I can't see any reason why you would object to using non-free software through a web interface. But if you think that it is not OK to use non-free software locally, then I have trouble understanding why you would think it was OK to use it through a web interface.
I can summarize my own position on this issue as follows:
If Machine A and Machine B are connected by an internet link and are doing non-trivial information exchange, they have become parts of a new machine. But users of the system are still individuals, and I think normal notions of freedom should continue to apply.
Here's a simple example illustrating a potential pitfall associated with web services that are non-free in the usual sense of the word. Take WebMathematica. I think someone operating a WebMathematica client will end up having just as bad a time as a user of regular Mathematica. WebMathematica and regular Mathematica are, in fact, the same program, just with different interfaces. Using this program when it is running elsewhere doesn't change matters for the better.
Google of course can't be "trivialized" in this way, because it can't be made to run locally. But, by the same token, one's freedoms as a Google user do not depend on what one can do on ones own machine; they depend on what one can or
I really think that if Google was free, it would not be so different from a normal software package, the fact that it needs to run on a centralized, special, expensive, machine in order to work notwithstanding. There would be package maintainers, filtered code-submission pipelines, user-level configurations, and so on; as it currently stands, there are none of these things (or, more precisely, they exist but aren't public), and I'd assume the Google product is the worse because of it.
The (late) "Gnugle" project talked about the creation of a free p2p search engine. Was this a pure fantasy? Or could something like that be made to work? Alternatively, if a large centralized system was needed, how could free software hackers ramp up to being able to afford such a system? Just how expensive would it be?
Note that the HDM project may have to do very significant search and processing operations on a corpus that may only be a few orders of magnitude smaller than the whole WWW. I am interested in developing software that can handle this sort of data intensity. But is something "like Google" really needed, or would a smaller supercomputing resource suffice? My crystal ball is clouded over. But anyway, hopefully this will partly explain where my interest is coming from.
I don't think this is a problem. In some cases, it would even be a good thing. This sort of contention is resolved by social means that have little or nothing to do with the licensing model. For example, CoopMart would have in situ modifiablity, and yes, there would be some struggle about who would control the individual pages. This struggle could also be called "public debate". The actual underlying software of a system like CoopMart (or any free system - Noosphere, Emacs, FreeSearch, whatever) would be maintained by people who know what they are doing and respond adequately to community feedback (including patches). In a case of total social meltdown, the projects could fork, after all!
I think there is a place for both free and non-free software. Same with web services. However I think the "legitimate" sphere of non-free services is larger than that of software. I think this has something to do with the service-vs-product dichotomy. With a service, you aren't "getting" an artifact, you're having some work done for you. Sometimes, as in Google, this is done by another party because you can't do it. But it could also apply if that party can do it more economically than you (by taking advantage of various economies of scale that come with applying the same service over and over again). In that case, a lot of aggregate resources might be wasted unless the closed implementation model is adopted.
Noosphere is a good example of a project that has nothing to lose by also releasing the source to the web service. However, we could hardly judge that a success as of yet, as contributions from those with whom I did not have personal connections have been negligible. Does this show that there simply is not the same open source productive dynamic in the case of web services? In that case, arguments about whether or not such software should be free are kind of academic. Perhaps payment for access to the web service (or the "torture" of having to look at ads) should be seen as punishment for situations where CBPP fails to produce the web service.
But, I will switch sides again by pointing out a reason why more web services should be free in the free-software sense: it would make them more resistant to control by concentrated interests. Just look at all of the bit torrent search engines. These monolithic web sites (which are first of all less useful for being fragmented) are easily shut down by Big Media interests, never mind that this is at best tantamount to banning a useful tool which might be used to commit a crime. Had these sites released their source, others would have been able to duplicate them, possibly in more resilient forms.
There is another twist that I think bears on this discussion. In a sense, things like the Bittorrent search engines and Noosphere aren't really web services, because they don't really have well-defined, machine-interpretable interfaces. This is actually what is meant when "web services" are referred to in internet architecture circles (in fact, the term can even be taken to mean a specific bundle of protocols). Part of the web services "movement" is to "open up" the web by having more of the previously human-only services mirrored by machine interfaces. Sometimes just putting some comments in HTML output or at least ensuring it is formatted cleanly and in a way that intentionally is not changed is enough to "open" a web service.
So, now we have a few levels of distinction:
These go from less to progressively more "free". Note that the distinction between the latter two arises precisely because of the service nature of web services: you can keep the underlying software closed, but in a sense make the system quite open anyway, provided the interface is transparent. Conversely, if you open the underlying system, you may be able to keep more of the interface opaque, since others may be able to set up services with customized versions of the interface.
But with the architecture of the internet and transparent web services, you can in an imporant sense "modify" a web service without ever touching the source of the implementation itself. This is actually analogous to scholia or superimposed information principles: you simply apply a transformation to the output of a web service before passing it along as your different/enhanced version. Utilizing this principle, a number of Google web service hacks have arisen, including the "Cheap Gas" Google Maps hack I showed up.
This principle is also the basis of much of the OCKHAM and MetaCombine? projects, where we are building web services that can be piped together with existing or new services.
This is all turning out to be very nuanced and interesting, but I should cut myself off for now =)
--akrowne Tue Jun 7 04:35:24 UTC 2005
I'd argue that even in cases with transparent interfaces, protocols, performance stats, etc., it is to the advantage of the public to see the code. For one thing, there's the education/portability factor: how does Google work? Are there algorithms that they use that I might be able to use as part of a theorem proving system? You've mentioned before that most of the "Google algorithm" is known, but it would be easiest if one could simply get it from a centralized place. Maybe I should stop whining about Google and just put together my own page giving a listing for their algorithm or the best approximation to it I can put together, together with discussion and ideas for related work. The point is that such a resource has value independent of whether or not it can be used/changed in situ. However, as we've been chatting about recently by email WRT something else, it is much more useful if it is free as opposed to simply just published. Thus, the code can be separated from the webservice to be an essentially unrelated good altogether. Or it can be a related good: you read the code and get some ideas about how to submit better queries, for example.
In situ modifiability is a whole 'nother can of worms. Maybe later we can talk more about that. The one (somewhat related) thing I want to say now is that I don't think we can conclude from the Noosphere experience so far that there "is not the same open source productive dynamic in the case of web services" - the only thing I can conclude about Noosphere is that it should be easier to install! We've definitely been over that. There are probably plenty of other things to say about it, but I'd need to know the code to say them.
--jcorneli Tue Jun 07 05:15:32 2005 UTC
I would say that this is the issue of "free as in beer" in another form. Web services which can be freely used but whose source is closed seem to me a lot like closed source "freeware". In addition to the points raised I would also point out that there is always the danger that the entity offering the product can always change its policy. For instance, if Mathematica considers that use of the free web version is cutting into sales of the desktop version, they might very well decide to start charging for it as well and noone could do much more than make faces and moan about it.
As for the web service hacks, they sound rather analagous to free Mathematica packages or free applications for non-free operating systems.
There is a sense in which the latter analogy might be rather exact. In Aaron's blog, there is an interesting suggestion that the web can be every bit as much a platform as an operting system. If this is so, then these would be a reason for concern. After all, GNU was founded exactly because of the issue of free operating systems and, if the web will complement or sometimes replace the operating system as a platform, it would be a pity if basic web services would not be available in free versions. Under such circumstances, the victory of Linux/GNU would be rather hollow.
I would like to point out that the fact that the Google search program runs on a computer of a size which few people have access to may not be very relevant. Remember that, as technology advances, features which were once only available on supercomputers tends to become standard feature of personal computers. For instance, two or three decades ago, only a supercomuputer would have come with a gigabyte of memory, a terabyte of disk storage, and gigahertz processing speed, but now these are the specifications of a typical laptop! It sems plausible that personal computers able to run the Google search program will come out in our lifetimes. However, given that copyright protection lasts for 95 years, it seems quite likely that, when such computers come out, it will only be possible to run the Google program on them with permission, likely by paying a licensing fee and that, by the time the copyright runs out, the work will be a historical curiosity.
I am not opposed to using and even promoting "free as in beer" content whether it be web services, computer applications, or electronic books providing that one makes it clear that, despite whatever the advertising hype may say, it is not really free. I worry about the possibility that people may be lulled into a false sense of security and be content with "good enough for government work". When someone is being a jerk about squeezing every last drop of profit out of their copyrights, it is easy enough to see the desirability of a free, open source alternative, but when one is playing the benevolent dictator by allowing free access to closed (or even open) source content (but jealously reserving their intellectual property rights and maybe even making up a few new ones of dubious legality), most people will not care about the issue of free intellectual content (or maybe just intellectual content which is they use free of charge because their university or some other such entity is footing the bill).
Finally, a crazy, silly thought just occurred to me. I wonder if Google would fund development of Gnugle as a project for its "Summer of Code" programme? --rspuzio 7 June 2005
Yes, if google runs on something 1000 times faster than today's PC, in maybe 2 decades we'll have PC's that can do what Google does if Moore's law continues to hold. Of course, by that time, Google and other computing supergiants will probably be on to even more interesting things that can only run on the supercomputers of 2025. I have trouble imagining what sort of amazing things these supercomputers would be able to do, but if they still have source code in the future, I'd it to be something I can see and modify. Maybe we'll all be living in The Matrix in 20 years, who knows, in which case user-level modifiability could give us important ways to hack our own lives and experiences. But if that's all we have and the underlying code is centrally controlled… well, you've seen the movie, it ain't pretty. Already there is something like this - the "social codes" that govern our behavior are often controlled by corporations and government, or at least we imagine them to be. WalMart and the local co-op provide no-cost (or low-cost) services: basically, the right to shop. But these things aren't really free as in freedom. We only have user-level freedoms: the freedom to drink Coke, Pepsi, or water.
When you guys talk about the "place for both free and non-free software" or being "not opposed to using and even promoting 'free as in beer' content", you aren't giving examples that say why non-free content is good. I'll try to help you out. I've seen "The Matrix" a couple of times. I've seen all the Star Wars movies. These things are proprietary and worth a lot of money. I currently am running Mac OS X. Even though I only use free user-level programs, I'm not a "software saint". And yet, when I think about the things that are most valuable to me, and the things I'd like to support in the future, I'm not at all convinced that proprietary systems have a place; so, if they do, what is it? Do we need copyright in order to have good movies to watch and interesting books to read? Do we need proprietary operating systems or end-user programs in order to do the sorts of computing we'd like to do? Do we need to have back offices where corporation-level decisions are made with "NO ENTRY" signs on the doors?
The answer could be "we need organization-level control for some systems" - clearly for families, for example. I don't think just anyone should be able to declare themself to be a part of my family and start changing the way the family works. In general, individual freedoms are important. But as systems become increasingly public - art, or internet, or shopping, or government, for example - I'm somewhat less convinced by a "freedom" argument. The freedom to obfuscate is rather different from the freedom to create. If WalMart pleads that it needs a "NO ENTRY" back room because otherwise a bunch of hippies might come in and change the way they do business - I'd probably say, damn straight (noting, of course, that co-ops have many of the same problems that WalMart has).
Anyway, "free software" doesn't mean "lack of control". These things are free but they aren't free-for-alls. All in all, I guess I'm having trouble making a case for non-free systems and services. Further input?