On 3/2/2012 8:37 AM, Martin Baldan wrote:
Julian,
I'm not sure I understand your proposal, but I do think what Google
does is not something trivial, straightforward or easy to automate. I
remember reading an article about Google's ranking strategy. IIRC,
they use the patterns of mutual linking between websites. So far, so
good. But then, when Google became popular, some companies started to
build link farms, to make themselves look more important to Google.
When Google finds out about this behavior, they kick the company to
the bottom of the index. I'm sure they have many secret automated
schemes to do this kind of thing, but it's essentially an arms race,
and it takes constant human attention. Local search is much less
problematic, but still you can end up with a huge pile of unstructured
data, or a huge bowl of linked spaghetti mess, so it may well make
sense to ask a third party for help to sort it out.
I don't think there's anything architecturally centralized about using
Google as a search engine, it's just a matter of popularity. You also
have Bing, Duckduckgo, whatever.
yeah.
the main thing Google does is scavenging and aggregating data.
and, they have done fairly well at it...
and they make money mostly via ads...
On the other hand, data storage and bandwidth are very centralized.
Dropbox, Google docs, iCloud, are all sympthoms of the fact that PC
operating systems were designed for local storage. I've been looking
at possible alternatives. There's distributed fault-tolerant network
filesystems like Xtreemfs (and even the Linux-based XtreemOS), or
Tahoe-LAFS (with object-capabilities!), or maybe a more P2P approach
such as Tribler (a tracker-free bittorrent), and for shared bandwidth
apparently there is a BittorrentLive (P2P streaming). But I don't know
how to put all that together into a usable computing experience. For
instance, squeak is a single file image, so I guess it can't benefit
from file-based capabilities, except if the objects were mapped to
files in some way. Oh, well, this is for another thread.
agreed.
just because I might want to have better internet file-systems, doesn't
necessarily mean I want all my data to be off on someones' server somewhere.
much more preferable would be if I could remotely access data stored on
my own computer.
the problem is that neither OS's nor networking hardware were really
designed for this:
broadband routers tend to assume by default that the network is being
used purely for pulling content off the internet, ...
at this point, it means convenience either requires some sort of central
server to pull data from, or bouncing off of such a server (sort of like
some sort of Reverse FTP, the computer holding the data connects to a
server, and in turn makes its data visible on said server, and other
computers connect to the server to access data stored on their PC,
probably with some file-proxy magic and mirroring and similar...).
technically, the above could be like a more "organized" version of a P2P
file-sharing system, and could instead focus more on sharing for
individuals (between their devices) or between groups. unlike with a
central server, it allows for much more storage space (one can easily
have TB of shared space, rather than worrying about several GB or
similar on some server somewhere).
nicer would be if it could offer a higher-performance alternative to a
Mercurial or GIT or similar style system or similar (rather than simply
being a raw shared filesystem).
better though would be if broadband routers and DNS worked in a way
which made it fairly trivial for pretty much any computer to be easily
accessible remotely, without having to jerk off with port-forwarding and
other things.
potentially, if/when the "last mile" internet migrates to IPv6, this
could help (as then presumably both NAT and dynamic IP addresses can
partly go away).
but, it is taking its time, and neither ISPs nor broadband routers seem
to yet support IPv6...
-Best
Martin
On Fri, Mar 2, 2012 at 6:54 AM, Julian Leviston<jul...@leviston.net> wrote:
Right you are. Centralised search seems a bit silly to me.
Take object orientedism and apply it to search and you get a thing where
each node searches itself when asked... apply this to a local-focussed
topology (ie spider web serch out) and utilise intelligent caching (so
search the localised caches first) and you get a better thing, no?
Why not do it like that? Or am I limited in my thinking about this?
Julian
On 02/03/2012, at 4:26 AM, David Barbour wrote:
_______________________________________________
fonc mailing list
fonc@vpri.org
http://vpri.org/mailman/listinfo/fonc
_______________________________________________
fonc mailing list
fonc@vpri.org
http://vpri.org/mailman/listinfo/fonc
