RMS issues ukase against Software as a Service – and I agree it’s an iffy idea

In a recent O’Reilly interview, Richard Stallman utters an anathema against software-as-a-service arrangements, calling them “non-free” and saying “you must not use it!” It would be easy to parody RMS’s style of uttering grave moralistic sonorities as though he were the Pope speaking ex-cathedra, but I’m going to resist the temptation because I think in this case his concerns are quite valid.

Richard says “If you do your computing on someone else’s server, you hand over control of your computing to whoever controls the server. It is like running binary-only software, only worse: it’s even harder for you to patch the program that’s running on someone else’s server than it is to patch a binary copy of a program running on your own computer.”

He’s quite right about this. The connotations of “free” and “non-free” are rather beside the real point, which is a pragmatic one. It can be risky to give up control of your data, and becoming dependent on a SaaS or cloud service is not something to do casually. I wouldn’t say you “must not”, that’s too moralistic for me, but I would say doing so is not wise or prudent.

In general, I don’t allow myself to rely on such services. And I have a rule: unless I can get all my data back through some sort of export or dump function in a non-obfuscated format, I won’t go there. I recommend this rule to others as well.

36 thoughts on “RMS issues ukase against Software as a Service – and I agree it’s an iffy idea

  1. >In general, I don’t allow myself to rely on such services. And I have a rule: unless I can get all my data back through some sort of export or dump function in a non-obfuscated format, I won’t go there. I recommend this rule to others as well.

    Does that include the times when the exported function is supplied by the open source community and not the provider itself?

  2. >Does that include the times when the exported function is supplied by the open source community and not the provider itself?

    Don’t see why it shouldn’t. Do you have an particular example in mind?

  3. But, but – with your G1 phone aren’t you using gmail/contacts/calendar?

    Or is that somehow different?

    I’m confused.

  4. >But, but – with your G1 phone aren’t you using gmail/contacts/calendar?

    Have ‘em. Don’t use ‘em.

  5. Basically, you’re talking about non-proprietary portability (transportability, exportability, etc.). Isn’t that pretty much the same standard by which you would choose any open source application? But what if the service did meet your criteria? Even if you could not (practically or actually) modify or patch the server functionality. Might you use it?

  6. >Basically, you’re talking about non-proprietary portability (transportability, exportability, etc.). Isn’t that pretty much the same standard by which you would choose any open source application?

    Yes.

    >But what if the service did meet your criteria? Even if you could not (practically or actually) modify or patch the server functionality. Might you use it?

    Maybe. I’d have to think carefully about the particular risks I was assuming.

  7. I trust Google apps more than other web apps because they do provide dump functionality and because I trust Google’s don’t-be-evil track record*. That said, there’s a different reason that I refuse to use GMail: I won’t entrust anyone with the ability to receive my email. Since every “lost password” link works through email, your inbox is basically a master key to your identity. I’m not risking that key to a rogue Googler. I run an SMTP server from my home system, and that’s the way it’s going to stay.

  8. Daniel F:

    Your inbox, as opposed to your home host security, or your DNS MX records?

    It always seemed to me public-key crypto was the answer to that problem, but either there’s a chicken-and-egg problem or (more likely) nobody much really cares, because it’s never taken off…

  9. Tying in with your previous post, the AGPL is the only thing that would protect us from Google going evil the way the GPL did against Microsoft. It’s quite telling that Google refuse to let AGPL projects onto Google Code.

  10. > Your inbox, as opposed to your home host security, or your DNS MX records?

    I’m not so paranoid as to think that someone at DynDNS would be specifically interested in my email and change my MX record in order to steal it. If they tried to do such a thing in bulk, it’d be quickly noticed. A Google employee, on the other hand, could just do a mass grep without having to interrupt any mail delivery.

    Host security is an orthogonal problem and applies equally regardless of who hosts my mail, but I have more confidence in my own ability to keep my box secure than I do in that of $GOOGLER[$RANDOM].

  11. It seems that Web-based services are the second-coming of timesharing. Could this resistance just be a reflexive aversion to time-sharing among old-time hackers?

    The obvious solution to the problem of exporting data is to set up Web services that export data in open formats.

  12. >Could this resistance just be a reflexive aversion to time-sharing among old-time hackers?

    Huh? We didn’t mind time-sharing. Actually, many of us rather liked it – there were strong microcommunities around time-sharing machines. I remember when networked workstations came in thinking I was going to miss that, and we didn’t get that sense of community back for many years – not until blogs and IRC and social-networking tools started to get good.

  13. >We didn’t mind time-sharing. Actually, many of us rather liked it – there were strong microcommunities around time-sharing machines.

    My reading of the various Jargon File entries seemed to indicate a preference for the greater control over one’s environment that PCs allow and the avoidance of all the various technical hiccups associated with timesharing. Would it be better to characterize the old-time hacker experience as a love-hate relationship?

  14. >Would it be better to characterize the old-time hacker experience as a love-hate relationship?

    Yes. It wasn’t time-sharing we minded, really, it was that the machines were so damned slow and other people were getting most of the clock cycles.

  15. Regarding gmail services: I think IMAP, POP, and ICAL access really take care of those issues – the address book is available for export as two different kind of csv files, as well as VCard.

  16. It’s worth mentioning that a lot of social networking services have RSS feeds, so periodically saving a local copy with a desktop RSS reader might be suitable workaround for the data-on-the-server problem. Most moral problems are really engineering problems in disguise.

  17. I remember back in the late 80s/early 90s when lots of people were getting free accounts on MIT’s lab computers, but I guess in that case it was OK to have your data on someone else’s computer because RMS was running the computers.

  18. rms was not an admin on the ITS machines “in the late 80s/early 90s”.

    He did work on ITS prior to Project GNU, but after he started that in early ’84, he ceased being active on the other lab machines.

    The last of the MIT ITS machines was shut-down in 1990.

  19. I have faith in natural selection. Even the best disaster recovery plan (typically) fails when a data center blows up (like The Planet did) and power cycles 20 times an hour for a day. Boom Boom, Out Go The Lights and said data center offers gift certificates for data recovery to its clients. So there goes your ‘cloud’ that could never fail. When I say fail, I mean that someone’s data can’t be accessed when its needed, I’m not implying that it can’t be restored later to some previous (or current) version. But then again, most start up companies that offer ‘foo 2.0′ live in a single facility.

    And then there is the blogger who discovers some kind of viral cookie that your service sets, ask Google about that. Then there is the (name 100 more things that separates the wheat from the chaff).

    Those who operate ethically and responsibly will flower, those who do not will wither. Its just the way things work. RMS is extremely paranoid, so I question his objectivity on the subject. I don’t want a lecture on my privacy, software freedom or anything else, I want examples of real technical issues.

    If the last 20 years have not taught people to back up their data locally and not believe everything they read on their screen .. well ..

    I don’t call people stupid, my faith in humanity lies in believing that if people knew better, they would do things differently.

    As time progresses, reliable and ethical services will stand out .. or perhaps none will and this fad will join the ranks of leisure suits.

  20. @esr re:Google. Sure, you an get the data for those services as you describe. But, just playing devil’s advocate here (I am a happy Gmail user), what happens if Google pulls a Yahoo circa Y2K era and locks those methods behind a pay wall? Without notice?

  21. >But, just playing devil’s advocate here (I am a happy Gmail user), what happens if Google pulls a Yahoo circa Y2K era and locks those methods behind a pay wall? Without notice?

    Well, of course. That’s why I don’t use gmail and don’t allow myself to rely on Google services that require them to store my data.

  22. or perhaps none will and this fad will join the ranks of leisure suits.

    God I hope so. “The cloud” is such a flawed metaphor for networked applications (much like “cyberspace” before it) that it has no place outside a meandering _why the lucky stiff Tweeter posting. (Except _why’s cloud is made of pink cotton candy and rains leprechaun gold…)

  23. @esr — My point is that if you’re relying on gmail, you need automated tools to retrieve your data for backup/offline use. An old project of yours, fetchmail, is one such tool. Nothing for Contacts or Calendar that I’m aware of — which is why I don’t rely on those. But maybe I could whip something up in Python… :)

  24. it has no place outside a meandering _why the lucky stiff Tweeter posting

    Do you mean http://www.djangopony.com and related stuff? This is not that meandering as you think, just a little bit of humor.

    Just curious, why do you think this metaphor is flawed?

  25. I mean _why the lucky stiff, the Ruby guy who frequently expresses himself with surrealist topic diversions and strange but cute cartoons.

    And I think the metaphor is flawed because it hides too many implementation details in its fuzzy abstraction. "Cloud computing" isn't really anything new, just another word for "networked applications". (Remember the "thin client" and "network computer" hype of the 1990s?) The term serves mainly to paper over the nitty-gritty details of implementation which do affect the end user: latency, outage, scalability, the fact that your freakin' data is on some remote server somewhere run by strangers, etc.

  26. I think he's the guy behind "Django pony" joke with candy pink clouds.

    My point was that sometimes seemingly surrealistic and strange things might have a deeper meaning.

  27. Which is why if _why starts talking about clouds, I'm going to pay more attention than when InfoWorld throws about the term. :)

  28. _why is obviously some deranged /. user and Ruby programmer who won’t let the “OMG! PONIES!” joke go, so he decided to resurrect it and make fun of Django at the same time.

    What he fails to realize is that I am soon to reveal my evil plot to unleash Ruby, the Rainbow-Loving Unicorn on Rails! Muahahahahah!

  29. @Morgan,

    You could use fetchmail, but Google uses imapsync to move your IMAP/POP mailboxes into gmail. I use it to keep a nearly-live backup @ home. Works fine.

    On the subject of Google calendars,

    Log into your Google calendar
    Click on the Manage Calendars link (On the left side)
    Click on the calendar you want to backup
    Scroll down to the section that says Private Address
    There will be a link that says “ical”, Thar e be yer data. matey!

    If you're on a Mac, you know what to do.
    If you're on a linux box, http://www.google.com/support/calendar/bin/answer

    All of this can be driven from the command line (or chron), too.

    Similar solutions are trivial for contacts backup.

    All this "I run my own servers" is just the bleating of the over-50 crowd, still fighting the war they lost.

  30. I don't use one. I use Gnus as my mail reader, and access mail from on the road using ssh and emacs 23's multi-tty mode.

  31. > All this "I run my own servers" is just the bleating of the over-50 crowd, still fighting the war they lost.

    Uh, I'm 23.

    (Sorry if this is a dupe; comment system is acting up.)

  32. I use gmail for one simple reason: I have access to my mail everywhere: am I at home in Moscow or am I at my girlfriend's place in St.Petersbourg or am I in US seeing my uncle — my mail is always with me. And one single cron script takes care of keeping all data where I want it. Older guys seems to be too conservative, sigh…

  33. I don't think so. google for "rb_parse_args" — _why is much better hacker then you think.

  34. >All this "I run my own servers" is just the bleating of the over-50 crowd, still fighting the war they lost.

    All else being equal, I would much prefer to run, and have all my data on, my own servers as opposed to using a service. I currently use Gmail, but I have no other choice except Comcast webmail, which is no better for your data (probably worse, because I trust Google more than Comcast) and is harder to use.

  35. In general, I don’t allow myself to rely on such services. And I have a rule: unless I can get all my data back through some sort of export or dump function in a non-obfuscated format, I won’t go there. I recommend this rule to others as well.

    yup. “me too” /AOL

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong> <pre lang="" line="" escaped="" highlight="">