31 replies to this topic

[littlesat]

 it's good and it works.

->

It's not good, but it does work.

 

Again.....

 

Why do you require NFS? Why is this really a must???

Edited by littlesat, 1 June 2016 - 15:20.

[gorski]

From my experience it is good and it works - otherwise this sounds like a dogma... Of course, you could explain, so we can see rationally what is the reasoning behind it...

 

NFS works better than CIFS in my long experience of these matters... And I do NOT know these things in depth, of course, being an IT lay person.

 

BUT! I can tell what works and what works better, that is for sure...

[WanWizard]

So if you don't know these things, how can you claim to know based on experience that it is good?

 

You're talking rubbish. Yes, their fix works. The same way as a tricycle works to move a mountain. They use a canon to kill a mosquito, and they haven't fixed the CIFS issue, instead the've implemented another canon.

 

We don't work this way. Code that goes into OpenPLi has to be of good standard, and should solve the problem. This is neither, and if you implement it, it will never be fixed properly. And over time, your entire codebase will become a mess, and you end up with an instable and slow codebase like OpenATV. One of the reasons people keep coming back to OpenPLi.

[malakudi]

You can always add the mounts manually, the network browser is just a helper for people too lazy to write the resource path (192.168.1.5:/mynas for example) with the remote. That's why the issue has low priority.

The oe-a fix is bad, because among other things it requires nmap. They were lazy too. A port scanner can be a few KBs of code, but they preferred to include nmap, a very large binary that has many many different functions just for utilising only a port scanner. And spawning an external binary just to do a port scan seems really BAD. Even if it works.

[littlesat]

@Malakudi,

 

Thanks for the complete description.... I was just preparing approx the same story....

[gorski]

I know, even though I am IT lay person, that in a small flash (older HW, like F500HD), this is a problem. But I did say that OpenPLi is better for older HW. I also said that in new HW, with larger flash and more firepower, this is not a problem. Did I not? Hardly rubbish. I also said I tested a lot (on various machines, various E2's, various versions) and I said I know from those tests what works, what doesn't work. Now, if you wanna be dogmatic - who's to stop you. But I know what I know. Period! No dogma needed! I have serious suspicions that this part of E2 can make E2 so seriously unstable, btw... Maybe I am wrong on this but...

 

So, sure: if you can sort it out in a better way - GREAT! And I mean it! Please do so! It would certainly be appreciated by plenty of us!

 

Having said that: I use OpenPLi based E2 in older HW. F500HD, Alien2 etc.

 

And CIFS works in OE-A. I don't care how. I do not use OpenATV but PBNIGMA, I also said that, have I not? OE-A based, also (not just OpenPLi). So, things are not that black/white, as you seem to be trying to portray them...

 

My point is: OK, if you have a problem, there is a solution which can be used, at least in the interim, until you manage to write a proper fix... At least in newer, stronger HW. That's all.

 

So, I think my suggestions were carefully put together, with some thought, qualified in a variety of ways...

Edited by gorski, 1 June 2016 - 19:42.

[lpnp]

I haven't got a chance to try it since I last posted.

The main reason I'm using NFS instead of CIFS is that the share is exported by a Linux host running on a weak multi-core ARM CPU, so I feel it's important to utilize all available resources to get as much performance as possible. NFS tends to perform better on Linux and it supports multithreading.

The only way I can think of to avoid scanning irrelevant hosts is to advertise the shares (e.g. mDNS/DNS-SD), but I don't think many users can be bothered to configure that.

As a compromise, is it possible to have a GUI where a user can specify a hostname/IP and get a list of shares on that host (showmount -e)? This can also be combined with avahi-browse/nmblookup to only list hosts with shares without having to port scan the whole subnet.

[gorski]

Don't you have it in the E2 GUI already, as I wrote?

[littlesat]

You can still connect NFS via the GUI.... only the network scanner does not find them...

[lpnp]

The network browser somehow worked, so I don't have to try other E2 workarounds.

automounts.xml also works fine with <ip>hostname</ip> . Network browser can show host names, why doesn't it use host name for mounts?

[Rob van der Does]

Not sure what the difference is, but for me NFS always show neatly in the networkbrowser:

 NWB-PLi-NFS.jpg   69.53KB   4 downloads

 

and can be used to create a mount.

 

Needless to say that the file 'exports' must exist on the targets

[WanWizard]

Correct.

 

The problem in the current network scanner code is limited to CIFS servers that have netbios switched off. Which is the default setting in modern Windows systems, which is why they won't be found.