Here's another monthly report to discuss the progress that has been made on Q-Filter over the past month. This update also gives some details on some of the problems experienced in during the past two days. We'll also cover some key changes and the direction the project will be taking moving forward.
After the transition over to the participatory rating system last month, scalability has been a concern to continue the growth of the system and web application. As more users begin to use Q-Filter, the home page begins to fill more and more and thus further methods of refinements were necessary to give more minnows the ability to showcase their work and gain visibility.
Thus, categories were a natural way to divide the posts into different areas of interest that a user might have. But rather than allow for any post with a tag belong to a category which Steemit and some of the other platforms do, the category restricts selection to only the primary tag. Thus as a result: A single post has a single category. This cleans some unrelated content from polluting the category which happens on Steemit.
Categories are organized on the site using a similar organizational scheme to Reddit. Subreddits use the /r/ structure. Categories on Q-Filter use the /cat/ structure. Pretty simple and easy to access. Categories use the same scoring mechanism as the home page which sorts all posts. Stake and Reputation scores are not available for categories for now.
To make things even easier to experience and get new users of the application familiar with categories, the Top 12 categories are now presented on the side menu for users to explore. These can change based on the activity in each category and some should provide interest to a wide variety of users.
Over the past 48 hours, the web application's interface has been timing out the home page due to the massive amount of calculations required to build each post's score. Given the increased traffic, the server was getting a pretty intensive workout. To improve service and speed up serving pages, the goal of nitro was to decouple the heavy calculations from the interface.
Thus, the calculations were removed from the service of the web page and moved into a background job. Thus, rather than calculate a new score every time the page is reloaded, a score is calculated continually every few minutes to update the scores of the posts.
There are three consequences of this action:
Webpages are now served 200 times faster than they were in the past. So you no longer have to wait several seconds for the web application to load and reload and the server no longer times out. It is better optimized for speed.
Refreshing no longer instantaneously updates the score. Thus you no longer have the ability to see a completely fresh score after you vote. This was the drawback of nitro and hopefully this can be worked on in the future. But you should see score that is fairly recent.
The way things were implemented meant that additional work would be required to implement this for stake-based and reputation-based algorithms. Since they were not implemented in categories and only available for the main feed, they have been temporarily removed for now.
Nitro's changes only effect the web application as the other systems were implemented independently of the web application. So if you use the Q-filter bot but not the web application, these changes do not effect you and you should expect service to run regularly.
The plan is to release a paper detailing the entire ecosystem and virtual network in detail. This basically serves to put all the information of the previous posts and place it in a single place. Also, more technical details will be provided for those interested. This is not a plan for a token release. Q-Filter is built on Steem and has no plans for now of going elsewhere or doing its own thing.
In the current updates, feeds have been removed. I would like to return those back to you and perhaps experiment with some new feeds. Also I have floated the idea of category customization, but it remains to see how such customization would be applied and who would be in charge of it on a category basis.
The current interface is ugly. Some improvement to modernize it would increase the appeal of the web application and prevent people from running away from its hideousness.
As mentioned in a previous post, the completion of basic categories marked the conclusion of the first phase of development in Q-Filter. While I will be working on the things above, I would also like to implement features that different users think are helpful.
Now would be a good time to point to the Q-Filter Github page here.
There's a nice area there to open issues which could be bugs or different proposals and I am sure to see them if you post your ideas there. You are also welcome to fork and play with the q-filter code. The application is written using Rails 5 so you'll need Ruby and a SQLite database to play around with the application.
You can also chat with me on Steem.chat. I go by the name 'greer184' there and would be happy to talk to about Q-Filter and any things you want to see implemented.
Lastly comments work as well, but I'm more likely and will probably be quicker through the other avenues, although I tend to be quick on Steemit as well. Thanks for reading and look forward to seeing you in next month's update or some of the quickie updates that occur in the meanwhile.
