>a) fishtools adds a Info.Format tag to the metadata of all its CHK@
>files (even when it's only used from a map file)
>
>b) FIW and fcptools use that small metadata you mentioned above
>
>c) some other tools may use even other metadata or use none.
>
>d) if the file has been dropped from freenet, you cannot insert it as
>you do not have it ;-) and if you have it, it has been inserted by you
>and so you can use your upload software to reupload it.
>
>When it goes to mapfiles, it seems to me that you can use the mapfile to
>determine the uploading client from it (from the order of the lines and
>the position of the default file).
>
>However, it would be good to have some "best current practice" for
>metadata attached to files...

definitive.. as long as there is "just a handfull" of insertion tools, there is still 
a chance to unite on a "common practice" or even create a definitive standard 
of which  metadata to include in CHKs.

>There are precisely two reasons. One is that it is a chunk in a
>splitfile, the other is that it is a file that has been redirected to by
>an SSK (possibly a manifest). These are two very similar cases. And
>there are tools which will let you do this. The trivial solution to get
>fproxy to do it would be to always create a redirect even if the file is
>a CHK, just to not include the metadata. However I am absolutely not
>going to do this, except maybe as an option. Because many sites DO
>insert files as CHKs and then link to them as 
>/CHK@<blah>/<preferred filename> And I see no pressing reason to
>implement such an option.

one should separate between the currently cruising CHKs with
- no metadata (data from SSKs or KSKs)
- small metadata (FIW, fcptools)
- more metadata (Fproxy)
- unknown metadata (the ones to come or the ones we just don't know)

and the future CHKs, as inserted as agreed on (please read on)

i don't want to loose backward-compatibility on existing CHKs so older sites won't 
work anymore, but rather define a common CHK metadata definition for 
future CHKs

>> d) if the file has been dropped from freenet, you cannot insert it as
>> you do not have it ;-) and if you have it, it has been inserted by you
>> and so you can use your upload software to reupload it.
>
>This ignores the case where you might want to "heal" someone else's
>edition-based site, when the original author has vanished (or has
>stopped reinserting it via empty transient nodes).
>
>Admittedly, that seems a somewhat unlikely case, but you never know.

sure, like you never know which will arise in the future

---~~---

now my view on CHKs:

CHKs are IMO pure "data holders". think of them as containers, which get a unique name 
on it because of the data which is held within.
so if you drop a text into one container, the name should be CHK@TEXT (TEXT = hash of 
text)

what you *do* with this text is not described, and it should not be described.
if you want to view it as text/plain or as text/html is of no concern for the data 
itself!
separate -content- and -view at the content-

example situation:

... so you uploaded your text into CHK@TEXT
and now you want to set up a "rule of how to view the plain data".
just insert a KSK@text-as-html, which consists of a metadata area, which liks to 
CHK@TEXT and describes the mimetype of text/html

if someone else has the same text or tries to reinsert your text, one must get a key 
collision, because the text *data* is already there
if someone tries to insert a "view CHK@TEXT as text/plain", then one will get NO key 
collision on the VIEW, but on the DATA

ending this scenary, you have three files:
CHK@TEXT --- the raw data
KSK@view-text-as-html  --- link to CHK@TEXT, and a mimetype of text/html, describing 
what to do with the TEXT
KSK@view-text-as-plain --- link to CHK@TEXT, and a mimetype of text/plain,describing 
what to do with the TEXT

if you add the view to the data, like the currently available programs do, you end up 
with just two files:
CHK@VIEW-TEXT-AS-HTML  --- text and "view as html"
CHK@VIEW-TEXT-AS-PLAIN  --- text and "view as plain"

smaller? i don't think so, because the data itself got doubled!

so cleaning up the relations between content and view (the data and the metadata 
section) will
- cleanse the metadata soup
- define a standard of metadata in CHKs (which should be NO METADATA AT ALL - IT'S 
DATA, NOT A VIEW AT IT!!!!)
- increase datastore capacity, because the data itself will collide more
- create a more flexible view-handling on data
- not allow questions to arise like "i have a mimetype specification in the map and 
one in the CHK, which one do i use?"
- make future insertion tools or other stuff consistent in metadata handling

thank you all for responding to my previous mail, i hope you will be equally engaged 
with this mail :)

HAND



_______________________________________________
devl mailing list
[EMAIL PROTECTED]
http://hawk.freenetproject.org:8080/cgi-bin/mailman/listinfo/devl

Reply via email to