[CollEc] CollEc 2.0 Access

Fri Jul 3 22:05:57 UTC 2020

Thanks for the messages.

Can you point test.collec.repec.org to http://95.216.35.87:8080?

I set up the user authentication to restrict access until the app is publicly released.

Start up speed varies greatly between sessions. Sometimes the app starts almost immediately and sometimes it takes quite a while or even fails. ShinyProxy handles this stage and I am not sure why it is so volatile.

I am currently running four parallel threads inserting data into the containerized MariaDB. This might at least partly explain the server's performance issues today. And it may adversely affect the app's loading times. The app containers are also attached to the same bridge network over which the data is inserted into the MariaDB - another potential reason for extended loading times.

The author search field is a selectize.js functionality. You can start typing at any position within an author's name. E.g. in my case you can start with "Christian", with "Düben" or even with "übe". You simply type something until the desired name shows up in the list of suggestions underneath. And those suggestions motivate the two reasons for which the app requires users to choose between names and handles before selecting an author. First, all available choices need to be loaded into memory and filtered while the user is typing. More choices make the functionality slightly slower. Second, a vector of names and handles combined also generates a list of suggestions containing both names and handles. And that might confuse users.

I am not sure what to write on where the screen is cached.

By "give people a list of top nodes at the start" you mean that the top default names on the list should be prominent, i.e. central, economists, right? I thought about it. And I think you have a point. I am going to change the default suggestions.

Many economists are not familiar with graph theory. And that is why the documentation explains the basics in a simply way. Edge weights are a key component of graphs and in my opinion one of the easiest parts to understand. Users do not have to be familiar with the slightly more complex shortest cost path algorithms or centrality measures to understand how edge weights work. If required, I can add another paragraph and a plot on edge weights to the documentation.

Cutting the introduction shorter is possible. Thanks for the suggestion. I will look into a few options.

Hosting the video on the server would indeed make the cookie notification obsolete. The motivation to embed a video hosted on Youtube comes from Youtube's video player. It automatically adjusts the resolution to the user's bandwidth, provides a rather stable performance, works well in different browsers and allows users to modify the speed. Other video players used online tend to be more annoying to use. They get stuck, complain about missing browser plugins, do sometimes not adjust dynamically to bandwidth limitations and come with fewer customizable settings. I can add the video via the HTML video tag. And then we check how it performs in the web application called by different browsers. If that does not work as intended we can still try other players or host it on Youtube.

I embedded the video in the introduction rather than the documentation because of feedback I received on GraphEc. The feedback was that users should understand right at the start how the app works - without having to click through the other tabs. The video is like a paper's abstract. It is a brief, convenient, initial overview presented to the user without him or her going through the details, i.e. the documentation. And besides that everyone likes to watch videos. Some users might not find it as enjoyable as funny cat videos, but it still makes the app appear more active and engaging from the start.

You mention the year 2013 in your e-mail. Should I use that as the year you introduced CollEc? I guess I somehow incorrectly understood CollEc to be a service from the early RePEc days.

The decision on whether your name shows up in the footer is up to you. It is your name. In the meantime I am going to check the footer's position.

Thanks for offering me to use the database on the host. However, I still prefer a containerized MariaDB. With a container I do not have to manually intervene when the database crashes. Docker simply spins up a new container with an uncorrupted MariaDB image without any loss of data.

A similar story holds for Java. I tried running ShinyProxy directly on the host. The required Java version was not compatible with the Debian installation. Therefore, ShinyProxy and the respective Java version are now deployed through Docker.

Shinyproxy runs in a container but exposes the app directly on the host's ports. So, would it be possible to set the A record to that port? Or to use a Nginx reverse proxy to link the port the URL is pointing to to the port ShinyProxy exposes the app on?

No worries about any delayed responses. That e-mail was meant as an update not invoking any feedback. And especially when you are sick you should focus on your health. Besides that, I am also busy with other matters, including teaching and ongoing research projects.

Let me make a few adjustments to the documentation based on your comments. Then I will send you the source code.

Have a nice day.

Christian Düben
Research Associate
Chair of Macroeconomics
Hamburg University
Von-Melle-Park 5, Room 3102
20146 Hamburg
Germany
+49 40 42838 1898
christian.dueben at uni-hamburg.de
http://www.christian-dueben.com

-----Original Message-----
From: Thomas Krichel <krichel at openlib.org> 
Sent: Freitag, 3. Juli 2020 15:41
To: Düben, Christian <Christian.Dueben at uni-hamburg.de>
Cc: CollEc Run <collec-run at lists.openlib.org>
Subject: Re: CollEc 2.0 Access

  Düben, Christian writes

> I am still working on the database performance. Insertion times 
> dropped but did not reach a satisfactory level yet. However, as the 
> application is done except for the database issue, I decided to seek 
> feedback from your side already. You can access it at 
> "http://95.216.35.87:8080"

  For that I already have test.collec.repec.org

> with the user name "Test_User"

  Why do I need to sign in?

  It's too slow after sign in. In general, darni is dreadfully slow today.
  I think the i/o is the bottleneck. I have been deleting files as we
  reached %94 full. 

  Few Otto Normaleconomists will understand the weighted networks.

  Data entry should not ask to fill a name or fill a handle,
  but detect whether the entry is a name or a handle. Ideally,
  one should enter family names as well as they are more famliar
  to users who search. 

  I think you need to start with a simple note, where the screen is cached.  

> The video embedded in the "Introduction" tab is a placeholder and will 
> be replaced with the actual tutorial upon the new CollEc's public 
> release. The output generated in the "Distances", "Closeness"
> and "Betweenness" tabs is temporarily not available as I am currently 
> working on the underlying MariaDB tables. I changed the documentation 
> a few times and am curious to hear your comments on it.

  I think it's better to give people a list of top nodes at the start,
  or just start with a screen very similar to the to one of
  the existing CollEc. It may not be a the best, but some folks
  got used to it. A video link should go to the docmentation
  page. 

-- 

  Cheers,

  Thomas Krichel                  http://openlib.org/home/krichel
                                              skype:thomaskrichel