Test automation "in Chinese"
Surely many have heard jokes about the Chinese. Who has not heard - I will give a couple of typical examples:
or this:
However, the tale is a lie, but there is a hint in it. Some time ago two interesting notes caught my eye:
This way of organizing work has been called crowdsourcing , and the range of its application is becoming wider. Our world is turning into a huge distributed superbiokompyuter, in which people play the role of processors. These are very good processors, they can solve such tasks that ordinary computers cannot yet do, and the role of traditional information systems comes down to providing communication between these human-processors.
And from that time on I was relentlessly tormented by the thought that automating certain types of testing was quite a suitable task for such a superbiocomputer. Functional testing, of course, will not be able to automate this way, because it is not easy to achieve good paralleling of a complex task on many processors. But there is load and configuration testing, there paralleling occurs in a natural way.
Yes, there are such services!
')
uTest offers services for functional and load testing applications. Litmus, which appeared a bit later, is aimed at website configuration testing.
I don’t know how Litmus really works, but this is how it seems to me. The site goes to testing, it is sent to the first, leading human processor, which in one browser, say, IE 7, passes through the site, builds and records chains, saves reference screenshots. After that, the site is distributed simultaneously to many other processors, each of which goes through the same chains and compares the observed result with the reference screenshots. At the same time, the ability of a person to fuzzy comparison of images is fully involved, because absolutely accurate pixel-by-pixel correspondence of displaying pages in different browsers cannot be achieved, and this is not necessary. Well, in the end, the results are collected from all processors and the client receives a general report. Voila! Superbiocomputer in action.
PS Crosspost
Chinese anti-tank platoon numbering 3000 people. consists of 3 offices for 1000 people. The task of separation - to disassemble the enemy's tank into parts, until he fired.
or this:
Today, a new powerful power plant has been commissioned in Shandun Province. Principle of action: three million Chinese are running on the ebony floor in wool socks.
However, the tale is a lie, but there is a hint in it. Some time ago two interesting notes caught my eye:
- Chinese search engine Baidu has launched a voice search service. Of course, the Chinese did not bother with the development of speech recognition technology. Everything is simpler - specially trained operators are sitting on that end, who accept the request, “recognize” it, perform a search and report the results to the customer either by SMS or by voice.
- Captcha Recognition. Recognition of the text of course the task is not as difficult as speech recognition. However, in captcha, the text is specially deformed and littered to complicate recognition. At the same time, it is simpler to invent and implement methods of littering than methods of recognition. However, captcha is still done so that a person can recognize the text, but the robot could not. Therefore, you just need to hire a lot of low-paid Chinese (well, or Hindus) who will sit and do recognition all day long.
This way of organizing work has been called crowdsourcing , and the range of its application is becoming wider. Our world is turning into a huge distributed superbiokompyuter, in which people play the role of processors. These are very good processors, they can solve such tasks that ordinary computers cannot yet do, and the role of traditional information systems comes down to providing communication between these human-processors.
And from that time on I was relentlessly tormented by the thought that automating certain types of testing was quite a suitable task for such a superbiocomputer. Functional testing, of course, will not be able to automate this way, because it is not easy to achieve good paralleling of a complex task on many processors. But there is load and configuration testing, there paralleling occurs in a natural way.
Yes, there are such services!
')
uTest offers services for functional and load testing applications. Litmus, which appeared a bit later, is aimed at website configuration testing.
I don’t know how Litmus really works, but this is how it seems to me. The site goes to testing, it is sent to the first, leading human processor, which in one browser, say, IE 7, passes through the site, builds and records chains, saves reference screenshots. After that, the site is distributed simultaneously to many other processors, each of which goes through the same chains and compares the observed result with the reference screenshots. At the same time, the ability of a person to fuzzy comparison of images is fully involved, because absolutely accurate pixel-by-pixel correspondence of displaying pages in different browsers cannot be achieved, and this is not necessary. Well, in the end, the results are collected from all processors and the client receives a general report. Voila! Superbiocomputer in action.
PS Crosspost
Source: https://habr.com/ru/post/56811/
All Articles