Re: [fonc] Unsolved problem in computer science? Fixing shortcuts.

2014-10-07 Thread Constantine Plotnikov
There are two assumptions that do not generally hold.
1) Both sides of links are always available at the same time (not true in
case of NFS that is often symlinked). Just consider the case of notebook
taken home and file system change in mean time.
2) There is a permission to change file system on the destination of the
link (not true in the case of readonly remote file systems and DVDs).
3) Also, why the server file system should absorb additional cost for the
each client? Is not it a hole to DoS it?

Best Regards,
Constantine Plotnikov

On Tue, Oct 7, 2014 at 8:41 PM, Casey Ransberger casey.obrie...@gmail.com
wrote:

 Context below, sorry about the top-post (stupid smartphone.)

 I think I remember that in Xanadu, links are two-way streets. When you
 move the link, I can only assume that both of those pointing devices
 would need to be updated.

 I'm not sure how it works though. Is there a central authority involved,
 can it be distributed, etc? It's hard to visualize a two-way link because I
 have spent my entire life living in flatland. I even mix up which plane is
 blue and which is pink sometimes:)

 The gist I got was that the two-way link concept was a powerful idea which
 could be applied to more problems than just pages (the mere use of the
 term is liable to give Ted a headache. Flat paper metaphors and such.) I
 wouldn't be shocked if a good implementation couldn't be done using a
 vigilant doubly-linked list (i.e. an object which cares about provenance
 and has a means of vetting it, like perhaps a touch of public key
 encryption.) Think of all the talk on this list about publish/subscribe as
 an object model, pattern directed invocation and such, and then try to
 imagine all of the ways a two-way link or shortcut might outclass the
 usual (and fragile-as-glass) one-way link.

 BCC Ted Nelson on the off chance that he might like to help us visualize
 the two-way link idea. (Ted, let me know if I shouldn't forward messages
 like this to you. Seems like giving some researchers a view into some of
 your ideas should help you on your way to realizing them. Then again, the
 road to hell is paved with... irritating people forwarding messages with
 good intentions.)

 Cheers,

 --Casey Ransberger

  On Oct 5, 2014, at 5:52 AM, John Carlson yottz...@gmail.com wrote:
 
  To put the problem in entirely file system terminology, What happens to
 a folder with shortcuts into it when you move the folder?   How does one
 automatically repoint the shortcuts?  Has this problem been solved in
 computer science?   On linux, the shortcuts would be symbolic links.
 
  I had a dream about smallstar when I was thinking about this.  The
 author was essentially asking me how to fix it.  He was showing me a
 hierarchy, then he moved part of the hierarchy into a subfolder and asked
 me how to automate it--especially the links to the original hierarchy.
 
  In language terms, this would be equivalent of refactoring a class which
 gets dropped down into an inner class.  This might be solved.  I'm not sure.
 
  This would be a great problem to solve on the web as well...does Xanadu
 do this?
 
  I think the solution is to maintain non-persistent nodes which are
 computed at access time, but I'm not entirely clear.
 
  I have no idea why I am posting this to cap-talk.   There may be some
 capability issues that I haven't thought of yet.  Or perhaps the capability
 folks have already solved this.
 
  For your consideration,
 
  John Carlson
  ___
  fonc mailing list
  fonc@vpri.org
  http://vpri.org/mailman/listinfo/fonc
 ___
 fonc mailing list
 fonc@vpri.org
 http://vpri.org/mailman/listinfo/fonc

___
fonc mailing list
fonc@vpri.org
http://vpri.org/mailman/listinfo/fonc


Re: [fonc] Unsolved problem in computer science? Fixing shortcuts.

2014-10-06 Thread Constantine Plotnikov
If you have symbolic reference, you have to be ready that destination is
gone, moved, or goes offline. Otherwise use hardlinks. You also could
create a defensive layer of symlinks just for indirection.

Nothing could be done in the distributed case anyway, because destination
could change while source of the link is offline.

Just to gerneralize things further, consider the case when renamed file is
referenced from other file. Should content of the referrer file to change
as well? After all symlink or shortcut is just a file that contains name of
other file as content, but treated specially by operating system.

BTW the problem is not a new one. The partinal solution for PL is called
refactoring.

Best Regards,
Constantine Plontikov

On Sun, Oct 5, 2014 at 4:52 PM, John Carlson yottz...@gmail.com wrote:

 To put the problem in entirely file system terminology, What happens to a
 folder with shortcuts into it when you move the folder?   How does one
 automatically repoint the shortcuts?  Has this problem been solved in
 computer science?   On linux, the shortcuts would be symbolic links.

 I had a dream about smallstar when I was thinking about this.  The author
 was essentially asking me how to fix it.  He was showing me a hierarchy,
 then he moved part of the hierarchy into a subfolder and asked me how to
 automate it--especially the links to the original hierarchy.

 In language terms, this would be equivalent of refactoring a class which
 gets dropped down into an inner class.  This might be solved.  I'm not sure.

 This would be a great problem to solve on the web as well...does Xanadu do
 this?

 I think the solution is to maintain non-persistent nodes which are
 computed at access time, but I'm not entirely clear.

 I have no idea why I am posting this to cap-talk.   There may be some
 capability issues that I haven't thought of yet.  Or perhaps the capability
 folks have already solved this.

 For your consideration,

 John Carlson

 ___
 fonc mailing list
 fonc@vpri.org
 http://vpri.org/mailman/listinfo/fonc


___
fonc mailing list
fonc@vpri.org
http://vpri.org/mailman/listinfo/fonc