Aiowch, my head hurts from this thread. Let's see if I can find the diametrically opposed inconsistencies and/or problems to shorten this thread:

1. limiting connections ala subscriber list vs. the matrix cafe/cell tower vs. DOS

2. Lack of a decent agent definition

3. Lack of clarity on the node an agent is running on as compared to accessing systems


Number 2 is a head scratcher. I think of it as an idiot-savante hacker or maybe an incredibly brilliant 6yro. Some simple personality features, a spoken-word interface that has trouble with particularly convoluted instructions, and a suite of stock skills.

I think I can handle #1. Matrix cafes, in the Starbucks-wifi-hotspot sense, do not exist since everyone is connected everywhere. Since most everyone has a comm, they really don't need to have rental computers for travelers. Which means a matrix cafe is actually more of a lan party, local-access only VR/AR environment. These places, like cell towers or even my house, use dedicated access points/routers that handle hundreds of connections for only a single subscription list entry.

Second, a residential DOS is based on bandwidth (SR4 unlimited). Servers can also be DOSed on CPU power (SR4 limited) but not actual connection overload. Individual comms won't be configured to run services so their Firewall will filter out all the junk requests. Port 440? Sorry, not here. Port 110? Get your mail elsewhere.

The Agent DOS can be handled by a frighteningly efficient, if logical, concept. Network blacklists. SR4 eliminates the network as being something you think about but it still exists and has hardware. If a DOS is detected the network (consisting of numerous repeaters, routers, etc) will begin communicating upstream to ID the Agent-laden node/host as DOS-zombies or Hackers, which should result in the node being blacklisted by virtually everyone with a hundred miles.

So not even Fuchi will consider a DOS on systems outside their perimeter if it results in their net access being revoked by the rest of the world.


#3 depends on how you view the persona software and agents. I need to re-read the RAW before I comment.

I thought that the subscription limit on limited the node as to how many other devices/nodes it could subscribe to. So if 100 people subscribe to the cafe's node then it is fine but if the cafe tries to subscribe to all 100 then there is a problem. But if one of those 100 (running a system 3 commlink) was trying to subscribe and already subscribed to 6 nodes he wouldn't be able to.

How about this to fix the subscription lists. You can have as many on your list as you want but you can't group them. That wil annoy the hell out of someone trying to do too many things. They have to take an action to deal with each and every drone they have.

Personally I don't see the point of this subscription list thing. Player laziness will eventually win out.

I think that there should be a Drone control program that can handle a certain number of drones. I think that Secretary programs (agents) should bring me virtual coffee, answer emails and do so in the nude in VR but not leave the commlink. I think that IC should be a figure of speech for the effect of a IDS smacking you.

I had an elequent and heart felt post about points of view but it got eaten by the system. Oh well.

Because Multiplexing is not hard to implement.

Anyway, Agents ordering Agantes are no problem.

Agents ordering agents are no problem? I'm going to asume that you are being sarcastic to be polite.

No. In fact, dasychaining nodes in a tree structure is the only way to build a security system by RAW.

How about this, guys.

An identical Agent is exactly capable of what any of it's identical bretheren are capable of, no more, no less.

If one Agent Smith fails in a task, they all will, because no Agent Smith is more innovative or clever than another. (And if one of them does manage to be more innovative or clever than the others, hire a Technomancer to permenantly delete that Smith!)

This is not to say that having more than one Agent working on one task is nessessarily a bad thing, or an untenable situation, far from it. But you need to buy a new Agent for each simultainous Agent you want working on one thing - Agents Jones, Johnson, and Stone, for examlpe.

Now, this dosen't hold true in Cybercombat with hackers versus Agents! Hackers are by nature innovative - but the same token is that they might not put up the perfect defense against Agent Smith's attack the next time!. So in Cybercombat, having a posse of Agent Smiths around to help you is a Good thing. But once one of them fall, the rest should be dispatched in short order, so it's still better to have a posse made up of Agent Smith, Agent Jones, Agent Johnson, Agent Stone, and Agent Orange, than it is to have 5x Smiths.

Smart people invented something called heuristic.

If your players are using Knowbots, I think it's pretty fair to say that balance went out the window a long time ago.

And that got what to do with Agents?
Any advanced decision-making involves heuristics, and thus is non-deterministic.
Agents make decisions, thus, they are based on heuristics... and thus won't come to the same results.

Otherwise, there is no point in rolling dice for an Agent.

Where this applies in the Agent Smith DOS issue is that all the Agents are going to be attacking through the Matrix connection, yes? Which really means they are attacking through the link between the Comm and the cell tower. Since the cell tower is the only device subscribed by the target comm, the Agents aren't going to overload the Comm that way. They would need to ID the target's GPS coordinates, identify multiple broadcast units in range of the Comm, gain access to said transceivers, and then begin to assault the Comm.

Yeah, that's when the network-level IC says "NO" and drops the blacklist in place.

DOS can't be hidden from the network as it depends on a wide number of connections made against a host. Stealth can't hide the connection request from the network because if the network doesn't see it, the connection doesn't exist and there is no DOS.

Network relays that transmit the existence of DOS further up the net will quickly kill DOS attacks. Spoof will keep your "real" comm account from being banned but won't keep the network from dropping your connection until you spoof another one. And while you are spoofing another one, the local network security kicks in, which means the wifi hub you are connected to goes offline or it refuses to allow connections to the DOS target. Plus, there's probably some kind of meat security called to make you stop.

DOS can be hidden from automated security. as long as the IC thinks that all of the incoming connections are legitimate, it won't do anything. however, the DOS attack will be immediately obvious to all the actual people who try to use the network, because they won't be able to. you can tart up a DOS all you want, but the basic purpose of the attack is something that is immediately apparent to everyone involved.

Hm, very useful discussion so far (though points tend to be argued over and over).
I think I know how to deal with DDoS and agent smith in my rules interpretation.

Furthermore I think Franks solution (the orchestrate multiple nodes idea) is very interesting and could make for a very interesting and simple rule set (because it is on a very high abstraction layer), but is conceptually very far from what RAW suggests (the lower node per node level) and thus would not be compatible with the spirit of RAW (while my rules are only contradict RAW but not its spirit).

At the moment I very interested wether Unwired will solve or ignore several matrix rules problems.

Given their record Serbitar I'm afraid that whatever is in Unwired will just makes things worse or just reinforce the problem areas.

Either that or just start writing, see what wonders you can come up with. Loads cheaper than using drugs to "get creative".

The one thing that has been running through my head over and over again in regards to all of these various 'broken issues' with hackers, agents and multiple commlinks, etc... is a single thought.

There is only ONE hacker trying to coordinate all this shit. Sure, he could have a million Agents working for him, but unless he wants to give them all identical orders, which is going to limit their utility, he can only command one at a time, and he's only got a finite amont of time to do stuff. Same with trying to overcome system limitations with mulitple commlinks taped together, he can only interface with one at a time, and changing which commlink he's using should take an action in it's own right. That's two seperate personas that have to hack in seperately, fight seperately, etc. Sure, SOME level of extra functionality could be gained by it, but given the physical limitations on the Hacker himself, it really puts a crimp in his style.


Now: Multiple agents given identical orders... some people suggest this isn't a problem for the hacker. No wonder their games have broken matrix rules. According to the RAW, agents are NOT AI's, they are reasonble facsimilies for interaction, but have a very very limited ability to function without orders. If you assume Pilot is equal to intelligence/logic/hacking then naturally you have a problem. It merely, in abstract, provides the same number of dice given equal ratings. Make your hacker actually write down his commands. For a million agents, telling them all to 'Hack that Firewall' is simple enough. But that's a million chances for the IC to notice the million identical attempts (assuming copied Agents, the only cost effective way to do it), and shut down to protect itself. So, you can effectively do a vandalizing DOS attack on that node. Peachy. Too bad not many Shadowruns are all about eliminating for 30 seconds use of a given node. YOu got to break in, not break it down.

This is where abstraction comes into play. It's not RAW, and I'm not claiming it is. Things like the million Agent Smiths are a reality of the world (in RAW it's possible, so in game, it would happen to any script kiddie hacker who broke copy protection on an agent program...). It's HOW people hack, or just one way. Your agent program doesn't represent a single Agent, it represents all the copies and copies copies that are slaved to it working from a single set of orders. This is why you have such limited numbers of active programs and subscription lists, because when you peel away the abstraction, they represent potentially dozens or even hundreds of programs running concurrently and all filling a single 'niche' represented by a single lable/dice pool. Exploit is all your exploit programs rolled into one rating, Firewall is all the firewalls you have daisychained that have to be defeated. Sometimes its just one awesome Chinese Military Icebreaker, othertimes is a suite of oldie but goodie hack warez you use like a master locksmith uses his picks. All fall under 'Exploit 5' or what have you.

Now, you could have two sets of a million Agents, that you use for two seperate sets of activity, thus you can have two 'AGENT' programs running from your Commlink remotely on other nodes. Again, the real limit is the exact orders they've been given, and the more complex the orders the more chances the players, and by extension the Characters, goofed up and left some idiot simple loophole out of the mix. Loaded Stealth, but didn't tell the Agent to use it. Loaded attack but didn't include commands to attack hostile IC. Sure, maybe the Agent is smart enought to eventually defend itself. Penalize it while it's internal 'instinct' programming kicks over by having it automatically 'surprised' when attacked. Whatever.

For dice rolling, sure the Agent(s) are still just as good as teh Hacker they replace, but for actual activities? Not so much. Again, I'm not worryign about RAW here, but if the Hacker was 'jumped in' to the Agent to try and circumvent the whole command line issue... guess what? He's using his Persona and commlink, not the agent anyway, the agent becomes fluff detail, something he replaced with his Persona when he took a more active hand.


Just my thought. We keep ignoring both the limits of the character's ability to interact with all this stuff (or handwaving it away...), and the GM's ability to say 'No, that's dumb'.

But... I'm tired and must go to bed. Night all...

@ Spike

I like your view on this up and think that I will start using it.

Just one problem with that, spike.

If you give players time to arrange their Agent's instructions like program code, they'll come up with the concept of having one or two or four "Monitoring" higher-level Agents whose only task is to monitor events and respond with preprogrammed orders - which involves giving multiple copies of identical Agents identical orders.

Example Monitor 1's only task in the whole wide web is to watch it's hacker's monitor for signs of his hacker being hit with Black IC. The very moment he gets simsense feedback, Monitor 1 orders Agent Smith 1 through Agent Smith 21 to follow the hacker's trail (which Monitor 1 of course knows, so they can get in without having to roll - they simply duplicate what Hacker 1 did that got him in the first time.)
The buck is immideately passed off when they get there, because Monitor 2 was waiting for the arrival of Smith 1 through 21. He already knows that if Smith 1 through 21 show up in his nizz-ode, the shit is going down, so he dosen't have to evaluate anything. He orders Smith 1 through 21 to attack the icon or icons in cybercombat with the hacker. They already know what and where the hacker is, they don't need to spend time Searching for him - and once you find him, which they do automatically, since they were preloaded with exactly that information (Location/ID) by Agent 1, they attack.

So if Black IC hits you, and you survive the first round, the second round your own posse of Agent Smiths show up to save you. And this is just a five-minute example.

So, the hacker stands there and gets hit for a few initiative passes like a punch drunk boxer on the ropes waiting for his cavalry to arrive? Not a good plan.

Pass 1, hacker gets hit for heavy damage, monitor one sees this
Pass 2 monitor sends the word for cavalry to come in, Black IC hit's hacker again
Pass 3 Cavalry mobilizes and GM forces them to make a hacking roll, because security (The IC) has already realized there is an intruder and closed that open channel. Black IC does the Cucaracha on the Hacker's medulla oblongata...
Pass 4 the Agent Cavalry pops in through the immedeatly aquires target on the black IC, which is busy pulling the Hackers metaphysical spleen through his nostril. The Black IC laughs.
Pass 5 agents rip Black IC apart as teh Sysop sees he's got a small army of invaders and shuts the server down. Meanwhile the hacker's skull, in real life, is smouldering nicely...

Maybe the agents get there two passes earlier... that hacker is still in terrible shape, if not actually dead.

if the hacker is dumb enough to go up against black IC that can kill him within a few passes, yes, that is what would happen. if, however, the hacker is not dumb (or, to be fair, unlucky), he will be squaring off against black IC that he can fight off for a few passes--long enough for the cavalrybots to arrive.

Even if the hacker isn't killed by the IC that fast... and to be honest, he probably didn't CHOSE to tangle with it, he is very likely to take physical damage that will take a decent amount of time to heal up and hinder him for some time.

Of course, the example I was responding to failed to take into account I also advocate folding mulitple iterations of the same program following the exact same orders into the abstract nature of the rules. It's only one agent coming to the rescue in rules terms, because any hacker worth his salt will be responding with waves of orders, codes and... yes... agents. All of which are represented by a single die roll. So you're 'Agent Cavalry' is probably not going to be a really effective rescue force, and are a terribly inefficent way to defend yourself in any regards.

If you're really paranoid, you go in with the Agent Cavalry with you.

That is, you and a single Agent sneak in. The Agent begins quietly creating more Agents, because that's his job - he's a Spawner, with specific orders.
The first thing he spawns is a Monitor, who's job it is simply to monitor the Hacker and respond based on preprogrammed variables. (Arguably, this function could also be carried out by the Spawner.)

Then he starts spawning Agents Smith, whose job it is to hurt hackers and break icons. This, they do rather effectively....

Really, my point is that it's better to just let Agents treat their pilot rating as an Intelligence rating. A Pilot rating of 6 means that they are in their element, and they've been programmed for pretty much every contingency. That way there's no need to set up elaborate and time-consuming (explaining time) schemes for sending in the cavalry.

So, you don't think the 'Agent Smith' thing is a bug, it's a feature. You WANT hackers to have armies of uberhacker minions at their beck and call in the matrix...

I'm down with that. Hackers become generals of virtual armies facing eachother down across the blasted wastelands of the Nodes, a millions agents a second Derezz every battle...

All of it abstracted to an attack program roll or a single agent roll, or what have you... that's not ONE black IC program trying to eat your brain, it's a swarm of them...

Unfortunately, the lack of option in matrix combat tend to make it less than interesting to run. Hopefully Unwired will make things more interesting, but until then I've often applied the melee modifiers to cybercombat. At that point the Many-to-One abstracting breaks down a little since people could start crying out for Friends in Cybercombat bonuses. Either way, I do like that abstraction, Spike.

To me the Agent Smith problem isn't a problem with combat. Sure if you want to take a few round of combat before the cavalry comes that's your perogative.

My problem with it arises when you have a million Agent Smith's doing a datasearch. Neo couldn't hide from Smith no matter where he was. Smith always found him. With millions of agents digging up dirt on everyone there'd be no secrets and that's not the essence of the game for me.

Here the anology breaks down - our Agent Smith is not The Agent Smith. It's not an AI like Deus.

It can only follow it's programming, complicated as that is.

Anyway, I'm not suggesting Agents should be virtual armies (though that would make for interesting Matrix imagery), I'm suggesting they should be hit squads.

Copies of the same Agent should get shut down easily, if you've shut one of them down. That's why you need to program different Agents from the ground up, to use different tactics and ideas. So if they crash your Smith, your Jones and Johnson are still there to pick up the slack.


Anyway, blatant Agent Abuse (Sending in armies) should be worth some kind of beat-down, I don't disagree. (Inless you go with the Hackers as Generals idea.) But intelligent use of one, two, or three Agent helpers? I have no problem with that.

According to the FAQ Agents can't load Agents.

I still don't see the reason for agents or most anything to actually travel around from node to node. They can access a lot of nodes without travel. It would be a resource nightmare to do this.

They describe agents as having a fairly good amount of decision making capabilities even if they are just a collection of decision trees. Everything I have heard about them makes them sounds fairly large. They are compared to Pilot programs that are operating as the OS for a drone.

So why is it better to expend the bandwidth to go in one direction at a time, moving from place to place, collecting information slowly instead of distributed packets paging multiple sites and waiting for the answer like we do today. A modern browser is more efficient then this.

Yes we will have larger bandwidths but why reinvent a system that is better and replace it with a bulkier system. It is like replacing the office intranet with a bunch of mail clerks. Think in terms of maximum effect from minimum action. Everything in the system can be written in realistic terms that would make modern sense with minimum traffic through the matrix.

You could have a series of VR CSS files that access default rooms in your reality filter to make one room look different from another. You can browse the internet just fine now. Just use that. You don't need the agents to do that. If you want a local secretary infomorph then I think that would make more sense then the hordes or mail clerks. If you really want cybercombat as the metaphor for your hacking then just realize that it is a reality filter on your computer and not actually what the system is doing. The matrix can have access to the effect of a bunch of tivos. This would probably replace normal television and cable access.

The Agent Smith Problem is a problem because you want it to be a problem. First off you are calling it Agent Smith which is half the problem right there. There is so many loaded assumptions in calling it that which you may not have but the person replying to your post does have. Too many people are basing the Shadowrun Matrix on a combination of Tron and The Matrix. I even think that the developers have fallen into that trap. Neither of them are reasonable models for this. One was a disney movie and the other was describing a prison.

Even if all of you come to a concensis then what? The developers don't really care. They already have your money. You can come to a discision for your game but if you are arguing for reality then you have a problem from the first step. The Shadowrun Matrix is just fantasy. It has left scifi behind because it is based on an illogical progression. I understand that the developers may have not known better in the beginning but they should have learned by now.

Now don't get me wrong and think that I am against house rules and different points of view. I just think that you trying to solve the problem of elephant poop with tape. Not many people are asking do we need this elephant here at all.

People should be asking how can we make this simplier and more reasonable for the players in the game. It is the players that take the brunt of abuse from our messing with the rules.

Do we need agents for the game to work? No
Do we need agents for the hacking rules to work? No
Do we need agents for browsing the matrix? No

We can make the agent into a secretary program and remove the problems AND confusion. It becomes just some fluff for role playing when a person accesses their commlink.

The agents aren't the point of the game and they have the possibility to move it away from the PCs. Therefore for that reason alone they should go.

You know, maybe we call it the Agent Smith problem because that's exactly how we see it?

Second, yes. The Shadowrun matrix is becoming more and more based off of The Matrix triology. This is a good thing.

Third, Agents are handy tools to have, and, if you Subscribe your mind to the School of Knasser, entirely managable.