zazukoians / trifid-ld Goto Github PK

This allows any scrip to know when rendering is finished.

The SPARQL proxy (/lib/sparql-proxy.js) should become a separate package.

We do provide VoID files for many data sets. Currently we did not define how they can be discovered for a resource.

This is handled in the VoID spec:

https://www.w3.org/TR/void/#discovery

So we need to have a way to support HTTP redirects per configuration.

SSL/TLS terminators like Hitch in combination with proxies like Varnish make it difficult to figure out for which destination port the request was made. This can be solved by using the PROXY protocol which is also supported by Hitch.

Proxywrap is a node-module which extends the Server interface of node to handle these situations properly.

Request I got:

Does Trifid have an icon i can insert in my slide to make some advertisement?

Based on discussions with @retog it makes sense to think about how we want to handle Backward and Forward links. Reto linked to a post from TBL in 2006.

We are not sure yet if we really always want to handle it that way so we need to discuss this a bit before implementation.

If we decide to implement this it will no longer work with DESCRIBE statements as we miss most of it. Reto implemented this functionality in Clerezza

The current configuration of port and host is confusing. Only the request object (default headers + proxy headers) should be used to build the IRI.

It's not top priority but I do think it would make sense to think about how we could integrate LD Fragments support as well. I probably could do it with Varnish and a LD Fragments server on another node instance but maybe there are smarter ways to do this.

Currently the predicates cannot be clicked in the RDF2h based output.

This is to allow loading scripts dynamically for example in the templates of RDF2h matchers.

There is no data on /resource/Eiffel_Tower on the html page, but the graph data is embedded in the html. In the configuration the dbpedia SPARQL endpoind is used, but the request goes to http://localhost:8080/resource/Eiffel_Tower. Here the example config.js.

To start it with iojs (using https://github.com/iojs/docker-iojs) I had to enter the following commands

npm install
iojs index.html

without the first it would complain about missing modules.

What to do after starting it is not clear, I would like to see what it can do using the default data. I figured that I had to replace localhost:8080 with the actual host and port in node_modules/tbbt-ld/dist/tbbt.nt. Now I can get some turtle at http://172.17.0.40:8080/data/person/stuart-bloom. But I can't access the simple HTML interface the readme mentions. Tried replacing localhost with 127.17.0.40 in config.js, but this didn't made any difference.

The handlers (/lib/sparql-handler.js, /lib/ldp-module-handler.js) should be moved to a separate package.

@retog could you please rebuild the docker image and push it to dockerhub? There is a change I need for SPARQL proxies.

Provide a generic way to list the contents of a namespace. Each path which is followed by a / shall trigger the following query:

CONSTRUCT {
  ?sub a ?obj.
}
WHERE {
  ?sub a ?obj.
  FILTER REGEX(STR(?sub), "^{?path}")
}

where the path might be something like

CONSTRUCT {
  ?sub a ?obj.
}
WHERE {
  ?sub a ?obj.
  FILTER REGEX(STR(?sub), "^http://classifications.data.admin.ch/municipality/")
}

We might add a LIMIT which should be fairly high, 10000?

The resources in the example dataset use port 8080. I.e., running trifid-ld on a different port breaks the example

I suppose the initial prefix list is handled in here for default HTML rendering. I noticed that schema.org is prefixed with s: in the default view. This is not aligned with RDFa initial context, see https://www.w3.org/2011/rdfa-context/rdfa-1.1

Please use all prefixes from there. I did some related work here already and I still think it would be a good idea if we would just use that for Zazuko stuff. Makes reading Turtle easier and less confusing with ambiguous prefixes.

It does not point to the correct endpoint yet by default. It should take the default SPARQL endpoint configured.

Currently one does not get any triples back from Trifid-LD when using rapper from Redland RDF Libraries. I filed an issue for rapper: http://bugs.librdf.org/mantis/view.php?id=590 (text taken from there):

Rapper currently (2.0.14) seems to send the following Accept header:

accept: 'application/rdf+xml, text/rdf;q=0.6, application/n-triples, text/plain;q=0.1, text/turtle, application/x-turtle, application/turtle, text/n3;q=0.3, text/rdf+n3;q=0.3, application/rdf+n3;q=0.3, application/x-trig, application/rss;q=0.8, application/rss+xml;q=0.8, text/rss;q=0.8, application/xml;q=0.3, text/xml;q=0.3, application/atom+xml;q=0.3, text/html;q=0.2, application/xhtml+xml;q=0.4, text/html;q=0.6, application/xhtml+xml;q=0.8, text/x-nquads, */*;q=0.1'

This works fine as long as RDF/XML is supported by the server. However, if the server only serves Turtle or NTriples, the server will answer to that with HTML before a RDF serialization.

RDF/XML is no longer supported by all servers out there so I would propose to give Turtle and NTriples a higher priority than HTML to make sure we get RDF back if it is supported by the server.

Dave answered with a question:

Can you be more specific.

Turtle / N3 family 0.1 to 1.0

application/n-triples, text/plain;q=0.1, text/turtle, application/x-turtle, application/turtle, text/n3;q=0.3, text/rdf+n3;q=0.3, application/rdf+n3;q=0.3, application/x-trig

HTML range 0.2-0.8

text/html;q=0.2, application/xhtml+xml;q=0.4, text/html;q=0.6, application/xhtml+xml;q=0.8

What would I raise and what lower?

Redland is still quite popular so I would like to know if we do something wrong or if Redland needs to adjust its request. I'm a bit lost here, maybe @retog could have a look at that and compare it with content-negotiation specs?

The contexts introduced in trifid templates should be withing its own namespace, and not squat existing namespaces as with "r2h:SameAsBlock

The patch headers middleware (/lib/patch-heders-middleware.js) should become a separate package.

Can we please rollback the UI stuff introduced in the last commit 74eb554? I do not want to introduce namespaces & names that will change and it does not yet provide the same functionality of the existing UI by @bergos.

This should either be removed or moved to a separate branch for the moment.

This is necessary so that they can be referenced in site-matchers.

Blocked by: rdf2h/ld2h#1

Verify meta-data visibility with https://search.google.com/structured-data/testing-tool/u/0/#url=http%3A%2F%2Fclassifications.data.admin.ch%2Fmunicipality%2F371

YASYGUI should be updated to latest version

For well known prefixes I would like to be able to show CURIEs instead of the full URI.

This requires a list of well known prefixes which I maintain here: https://github.com/uduvudu/uv-context.

Note that services like prefix.cc are ambiguous, that's why I started maintaining this list.

It would be nice if one can start without specific config file for a SPARQL endpoint.

Probably needs

• --sparql-endpoint-url
• --dataset-base-url
• --port

There should be a package to patch the request object to redirect requests to a different IRI.

The rendering of blank nodes looks pretty weird, see the attached screenshot

Sample URI: http://environment.data.admin.ch/ubd/28/qb/dsd_ubd28

• the http://purl.org/linked-data/cube#order attribute only shows up on the first one, but it's attached to all of them
• the int-value of this attribute is directly glued to the predicate
• the alignment is broken, looks different all the time

I just changed a config to listen to port 8080 instead of 80. i redirected the port so that externally it was still visible under port 80, nevertheless the resource trifid tried to dereference was http://hostname:8080/path rather than the actually accessed http://hostname/path. I think trifid should not try to be smart and in the absence of specific rewriting config assumed the requested URI should be used.

There is revision tagged as 0.7.0 which doesn't match the version in package.json. Versions should be changed with npm version to ensure things stay in sync.

I suggest now to update to 0.7.1 to ensure npm and docker are in sync again.

It seems that Umlauts are still broken in current release, see for example:

http://environment.data.admin.ch/ubd/28/qb/ubd28

The readme should contain a section about the configuration of reverse proxies.

We are still on a very old node image, update Docker to 6.4.x tags

https://hub.docker.com/_/node/

The HTML file for YASGUI contains hardcoded the endpoint URL. With a template, the server could replace the string with an URL from the configration. First it should check the SparqlProxy settings, than the SparqlHandler settings:

config.sparqlProxy.options.endpointUrl || config.handlerOptions.endpointUrl

This would allow for example to link to the RDF version from the HTML representation.

We should switch to the YASGUI 2.x tree after testing.

The default template doesn't contain any div block in the body which could be used to extend the template. A div block with a specific should be added an the JS code should be changed to use that div block.

So it looks more generic

Given this config for Trifid developer branch:

{
  "baseConfig": "trifid:config-sparql.json",
  "sparqlEndpointUrl": "https://web-sparql-virtuoso.dev.bgdi.ch/sparql",
  "datasetBaseUrl": "https://ld.geo.admin.ch/",
  "staticFiles": {
    "alodFiles": {
      "path": "/",
      "folder": "cwd:data/public"
    }
  }  
}

And this URI for testing: http://geotrifid:8080/boundaries/cantons/12 (localhost should probably work as well)

Trifid does the ASK for http, not https as it should according to the datasetBaseUrl.

It should be possible to embed trifid-ld in any node.js application, the embedding application should be able to provide a configuration as well as an overlay dir to override any file in data/public or to provide additional files to be served. This feature should be documented.

Currently trifid assumes that the schema is HTTP when reconstructing the request URI so it cannot be used if it exposed to the outside via HTTPS-URIs. As afaik there is no way to detect the schema of URI from the request headers wether to assume HTTP or HTTPS should be a setting in config.js.

By default config.js should be used, which should not be part of the default Trifid-LD repository. If that file doesn’t exist config.default.js should be used.

LDApp implements this already.

See https://en.wikipedia.org/wiki/HTTP/1.1_Upgrade_header

It's quite an obscure and probably rarely implemented header, essentially for changing protocol (request initiated by client). But I think it could potentially be very useful e.g. for shifting to a streaming protocol for media or some kind of fast-but-unreliable data transfer via UDP - triple per packet..?

I'm quite interesting in playing with it from a RESTful agents kind of perspective, but please consider this low priority.

(I haven't checked if OPTIONS and PATCH are supported yet - they'd both be much higher priority)

AntJS should be used to load all modules (logger, headers patcher, SPARQL proxy, search, …).

LDApp implements this already.

• dataset-base-url needs a / at the end
• keep url as a term but explain why not URI

Useful for checking e.g. if a suitable RDF representation is available at a given URL.

Minimal version would be something along the lines of :

var http = require('http');

var options = {method: 'HEAD', ...

var req = http.request(options, function(res) {
echoToClient(JSON.stringify(res.headers));
}

The readme says:

Note that it is not recommended to run Node applications on well-known ports (< 1024).

Shouldn't this speak about "privileged" rather than "well-known" ports.

http://www.w3.org/Daemon/User/Installation/PrivilegedPorts.html