[jbionic]

Hi there,

I am curious about one particular aspect of eMule. Unlike Torrents-like P2P, which rely heavily on external middlemen-trackers, eMule has an in-built search engine.

Now lets say I have a bunch of photos (primarily in *.jpg or *.png ) metatagged by IPTC or MWG with the use of ExifTool. The content of my collection can be sensitive to copyrights claims. Imagine that I wanna share my collection with someone else, so a person could search my collection remotely by metatags and retrieve the images without disclosing neither mine, no its own identity. Is this something eMule can help to facilitate?

Thanks in advance for guidance.

This post has been edited by jbionic: 23 November 2022 - 03:06 PM

[emule_user_downunder]

jbionic, on 24 November 2022 - 01:03 AM, said:

Hi there,

I am curious about one particular aspect of eMule. Unlike Torrents-like P2P, which rely heavily on external middlemen-trackers, eMule has an in-built search engine.

Now lets say I have a bunch of photos (primarily in *.jpg or *.png ) metatagged by IPTC or MWG with the use of ExifTool. The content of my collection can be sensitive to copyrights claims. Imagine that I wanna share my collection with someone else, so a person could search my collection remotely by metatags and retrieve the images without disclosing neither mine, no its own identity. Is this something eMule can help to facilitate?

Thanks in advance for guidance.

That would involve a whole lot of additional code to examine each file metadata and store it, for it to be searchable. You would have to extend the protocol and then interoperatability would suffer. One of the eDonkey variants might want to explore this however.

[jbionic]

emule_user_downunder, on 24 November 2022 - 07:39 AM, said:

That would involve a whole lot of additional code to examine each file metadata and store it, for it to be searchable. You would have to extend the protocol and then interoperatability would suffer. One of the eDonkey variants might want to explore this however.

Thanks. And that is still not the worst part of it. Metadata can be subject to frequent changes (tags updates and deletions), which results in re-calculations of hash values for each file.

There is also problem with fake metadata (a picture portraying a pet can be tagged as "lion" or "Jupiter"), so in order to ensure the validity of the shared content users would need to have some sort of rating system (perhaps separate ratings for owners and each piece of their collection too). While some a*s-h*les may still spoil it by upvoting the fakes once in a while. So unless image recognition and tag assignment are fully automated (by using advanced OCR engines and machine-learning) I doubt we'll see an effective solution any time soon.

There have been literaly dozens and dozens of publications in the last 20 years, if one googles by keywords "Searching in peer-to-peer networks" or "Content-Based Image Retrieval", but they still bring us nowhere.

[emule_user_downunder]

A job for artificial intelligence. Something that would be out of the scope of a file sharing software package to bundle.

[slivovice]

Comment in emule source:
"wrong meta data is driving me crazy because wrong meta data is even worse than missing meta data."

[jbionic]

There is more buzz on the forums about the need for collaboration tools for picture sharing and knowledge-organisation.
If you check the latest questions asked in the software recommendations community of stackexchange, there is a particular tag called "knowledge-organization" that seems to highlight the problem (unfortunately I am still unable to share links on the forum to point one)


There is a tool developed by Richard Eigenmann that claims to have the in-built picture sharing functionality, it's called Java Picture Organizer
- but I don't think the tool supports the possibility of running remote search queries based on meta-tags.

Surely, it'd be nice to have an app that would be able to fetch pics based of their meta-data. As pointed above, the crucial part of its workability would be trust.

So on one hand, the app must be resistant to copyright claims because of the legal nature of inforgraphic materials that people often generate and share (so the relevant web service must be somewhat remindful of LibGen). Thus privacy must be the cornerstone.

On the other hand, due to the fact that meta-data can be easily faked, the most relevant network topology for such an app must be F2F rather than P2P. While the list of known F2F solutions is rather short: Retroshare, WASTE, GNUnet, Freenet and OneSwarm. It must be a community-driven tool for small groups of collaborators who would know each other and who would have no interest in tampering with meta-data to falsify it (perhaps IPFS-cluster, or so called Private IPFS will fit the requirements too)

This post has been edited by jbionic: 31 March 2024 - 03:09 PM